Semi-supervised semantic segmentation (S4) has advanced remote sensing (RS) analysis by leveraging unlabeled data through pseudo-labeling and consistency learning. However, existing S4 studies often rely on small-scale datasets and models, limiting their practical applicability. To address this, we propose S5, the first scalable framework for semi-supervised semantic segmentation in RS, which unlocks the potential of vast unlabeled Earth observation data typically underutilized due to costly pixel-level annotations. Built upon existing large-scale RS datasets, S5 introduces a data selection strategy that integrates entropy-based filtering and diversity expansion, resulting in the RS4P-1M dataset. Using this dataset, we systematically scale up S4 into a new pretraining paradigm, S4 pre-training (S4P), to pretrain RS foundation models (RSFMs) of varying sizes on this extensive corpus, significantly boosting their performance on land cover segmentation and object detection tasks. Furthermore, during fine-tuning, we incorporate a Mixture-of-Experts (MoE)-based multi-dataset fine-tuning approach, which enables efficient adaptation to multiple RS benchmarks with fewer parameters. This approach improves the generalization and versatility of RSFMs across diverse RS benchmarks. The resulting RSFMs achieve state-of-the-art performance across all benchmarks, underscoring the viability of scaling semi-supervised learning for RS applications.

Jiawei Zhou, Chi Zhang, Xiang Feng, Qiming Zhang, Haibo Qiu, Lihuo He, Dengpan Ye, Xinbo Gao, Jing Zhang. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

Recommender systems traditionally represent items using unique identifiers (ItemIDs), but this approach struggles with large, dynamic item corpora and sparse long-tail data, limiting scalability and generalization. Semantic IDs, derived from multimodal content such as text and images, offer a promising alternative by mapping items into a shared semantic space, enabling knowledge transfer and improving recommendations for new or rare items. However, existing methods face two key challenges: (1) balancing cross-modal synergy with modality-specific uniqueness, and (2) bridging the semantic-behavioral gap, where semantic representations may misalign with actual user preferences. To address these challenges, we propose Multimodal Mixture-of-Quantization (MMQ), a two-stage framework that trains a novel multimodal tokenizer. First, a shared-specific tokenizer leverages a multi-expert architecture with modality-specific and modality-shared experts, using orthogonal regularization to capture comprehensive multimodal information. Second, behavior-aware fine-tuning dynamically adapts semantic IDs to downstream recommendation objectives while preserving modality information through a multimodal reconstruction loss. Extensive offline experiments and online A/B tests demonstrate that MMQ effectively unifies multimodal synergy, specificity, and behavioral adaptation, providing a scalable and versatile solution for both generative retrieval and discriminative ranking tasks.

Abstract In recent decades, visual simultaneous localization and mapping (vSLAM) has gained significant interest in both academia and industry. It estimates camera motion and reconstructs the environment concurrently using visual sensors on a moving robot. However, conventional cameras are limited by hardware, including motion blur and low dynamic range, which can negatively impact performance in challenging scenarios like high-speed motion and high dynamic range illumination. Recent studies have demonstrated that event cameras, a new type of bio-inspired visual sensor, offer advantages such as high temporal resolution, dynamic range, low power consumption, and low latency. This paper presents a timely and comprehensive review of event-based vSLAM algorithms that exploit the benefits of asynchronous and irregular event streams for localization and mapping tasks. The review covers the working principle of event cameras and various event representations for preprocessing event data. It also categorizes event-based vSLAM methods into four main categories: feature-based, direct, motion-compensation, and deep learning methods, with detailed discussions and practical guidance for each approach. Furthermore, the paper evaluates the state-of-the-art methods on various benchmarks, highlighting current challenges and future opportunities in this emerging research area. A public repository will be maintained to keep track of the rapid developments in this field at https://github.com/kun150kun/ESLAM-survey .

Due to the substantial domain gaps in Remote Sensing (RS) images that are characterized by variabilities such as location, wavelength, and sensor type, Remote Sensing Domain Generalization (RSDG) has emerged as a critical and valuable research frontier, focusing on developing models that generalize effectively across diverse scenarios. However, research in this area remains underexplored: (1) Current cross-domain methods primarily focus on Domain Adaptation (DA), which adapts models to predefined domains rather than to unseen ones; (2) Few studies target the RSDG issue, especially for semantic segmentation tasks. Existing related models are developed for specific unknown domains, struggling with issues of underfitting on other unseen scenarios; (3) Existing RS foundation models tend to prioritize in-domain performance over cross-domain generalization. To this end, we introduce the first vision foundation model for RSDG semantic segmentation, CrossEarth. CrossEarth demonstrates strong cross-domain generalization through a specially designed data-level Earth-Style Injection pipeline and a model-level Multi-Task Training pipeline. In addition, for the semantic segmentation task, we have curated an RSDG benchmark comprising 32 semantic segmentation scenarios across various regions, spectral bands, platforms, and climates, providing comprehensive evaluations of the generalizability of future RSDG models. Extensive experiments on this collection demonstrate the superiority of CrossEarth over existing state-of-the-art methods.

Video-text retrieval aims at returning the most semantically relevant videos given a textual query, which is a thriving topic in both computer vision and natural language processing communities. This article focuses on a more challenging task, i.e., Unsupervised Domain Adaptation Video-text Retrieval (UDAVR), wherein training and testing data come from different distributions. Previous approaches are mostly derived from classification-based domain adaptation methods, which are neither multi-modal nor suitable for retrieval tasks. They merely alleviate the domain shift while overlooking the pairwise misalignment issue in the target domain, i.e., there exist no semantic relationships between target videos and texts. While Foundation Models like CLIP perform well in in-domain video-text retrieval, their effectiveness significantly drops during domain shifts due to this lack of alignment. To tackle this, we propose a novel method named D ual A lignment D omain A daptation ( DADA ++). Specifically, we first introduce cross-modal semantic embedding to generate discriminative source features in a joint embedding space. Besides, we utilize cross-modal domain adaptations to balance the minimization of domain shift in a smooth manner. Furthermore, we empirically identify the pairwise misalignment in the target domain, and thus propose the i ntegrated D ual A lignment C onsistency (iDAC). The proposed iDAC adaptively aligns the video-text pairs, which are more likely to be relevant in the target domain, by verifying their cross-modal semantic proximity reciprocally in both hard and soft manners. This enables positive pairs to increase progressively while potentially aligning noisy pairs throughout the training procedure. We also provide insights into the functionality of DADA ++ through the lens of Foundation Models, explaining its superiority in a theoretical way. Compared with state-of-the-art methods, DADA ++ achieves 9.4% and 8.5% relative improvements on R@1 under the settings of TGIF \(\rightarrow\) MSR-VTT and TGIF \(\rightarrow\) MSVD, respectively, demonstrating its superior performance.

3D single object tracking (3D SOT) in LiDAR point clouds is essential for autonomous driving. Most existing 3D SOT methods focus on clear weather, where point clouds are more defined. However, adverse weather conditions lead to sparser and noisier point clouds, significantly degrading tracking performance and posing safety risks. In this study, we introduce UAWTrack, a universal 3D SOT model designed to perform effectively across diverse real-world weather conditions. UAWTrack comprises three key modules: 1) Voxel Feature Extraction, which mitigates the perturbations in point clouds caused by adverse weather; 2) Motion-centric Spatial-temporal Aggregation and Motion-guided Feature Fusion, capturing motion clues and sampling dense BEV motion features to address the issue of sparsity; and 3) Weather-Specific Tracker, which efficiently handles tracking in various weather conditions. To fill the gap of lacking benchmarks for 3D SOT in adverse weather, we simulate physically valid adverse weather conditions on the KITTI and NuScenes datasets, creating two benchmarks: KITTI-A and NuScenes-A. Extensive experiments demonstrate that UAWTrack achieves state-of-the-art performance under all weather conditions.

In the field of Moving Infrared Small Target Detection (MIRSTD), current methods typically use sequential modeling with two individual modules for spatial and temporal processing. However, such a modeling strategy lacks clear guidance on the motion and displacement difference between moving targets and background noise, thereby limiting the feature discriminability and resulting in error-prone target localization. This paper addresses this issue from clip and frame levels and proposes a novel architecture MOCID for MIRSTD. For clip-level feature fusion, we design a spatio-temporal backbone consisting of several proposed Fourier-inspired Spatio-temporal Attention (FISTA) layers. Each FISTA layer sequentially processes the features from spatial and temporal views to capture clip-level temporal motion context, where Fourier Transformation and Inverse Fourier Transformation are employed for each view. This context is then embedded into dynamic convolutional kernels for subsequent spatial feature extraction, thereby enabling clear motion difference guidance and generating comprehensive features. For frame-level feature fusion, we design a Displacement-aware Mamba Module (DAM) to capture detailed frame-to-frame displacement information. DAM utilizes an innovative Temporal Interpolation and Displacement-aware Scan technique to perform spatio-temporal difference-aware displacement modeling, introducing elaborate temporal indicators into feature extraction. Combining the above improvements, our model captures comprehensive motion and displacement contexts, significantly improving the detection of the small target. Extensive experiments demonstrate that MOCID achieves state-of-the-art detection accuracy on popular IRDST and DAUB datasets. Furthermore, MOCID offers a superior balance between throughput and performance compared to other methods. The code for this work will be made publicly available.

Single-frame Infrared Small Target (SIRST) detection has made significant advancements, but it still faces challenges due to limited labeled data and the foreground-background class imbalance. To address these issues, we introduce a novel Semi-Supervised SIRST Detection (S^3D) pipeline in this paper. First, drawing inspiration from thermodynamics, we propose augmenting infrared images using both chromatically and spatially uneven perturbations. This dual-stream perturbation enhances the diversity and balance of infrared samples, contributing to the robustness of detection models. Additionally, we develop a confidence-adaptive matching method to maintain weighted consistency among perturbed unlabeled samples. Second, to tackle class imbalance in labeled data, we compel the model to generate discriminative predictions for challenging, misclassified examples while down-weighting well-classified examples. We achieve this by modifying the standard cross-entropy loss to squeeze the detector and truncating the loss on well-classified examples. Our innovative Truncated Squeeze (TS) loss focuses on learning discriminative representations for difficult cases and prevents over-optimization for simpler ones. To assess the effectiveness of the perturbation techniques and loss functions, we apply them to various SIRST detectors and conduct comprehensive experiments on two benchmark datasets. Notably, our proposed methods consistently and significantly improve accuracy. Remarkably, our approach achieves over 98% performance of the state-of-the-art fully-supervised method using only 1/8 of the labeled samples.

Lingfeng Zhang, Xiaoshuai Hao, Qinwen Xu, Qiang Zhang, Xinyao Zhang, Pengwei Wang, Jing Zhang, Zhongyuan Wang, Shanghang Zhang, Renjing Xu. Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2025.

The astonishing breakthrough of multimodal large language models (MLLMs) has necessitated new benchmarks to quantitatively assess their capabilities, reveal their limitations, and indicate future research directions. However, this is challenging in the context of remote sensing (RS), since the imagery features ultra-high resolution that incorporates extremely complex semantic relationships. Existing benchmarks usually adopt notably smaller image sizes than real-world RS scenarios, suffer from limited annotation quality, and consider insufficient dimensions of evaluation. To address these issues, we present XLRS-Bench: a comprehensive benchmark for evaluating the perception and reasoning capabilities of MLLMs in ultra-high-resolution RS scenarios. XLRS-Bench boasts the largest average image size (8500×8500) observed thus far, with all evaluation samples meticulously annotated manually, assisted by a novel semi-automatic captioner on ultra-high-resolution RS images. On top of the XLRS-Bench, 16 sub-tasks are defined to evaluate MLLMs’ 10 kinds of perceptual capabilities and 6 kinds of reasoning capabilities, with a primary emphasis on advanced cognitive processes that facilitate real-world decision-making and the capture of spatiotemporal changes. The results of both general and RS-focused MLLMs on XLRS-Bench indicate that further efforts are needed for real-world RS applications. We have open-sourced XLRS-Bench to support further research in developing more powerful MLLMs for remote sensing.

Infrared Small Target Detection (IRSTD) aims to identify low signal-to-noise ratio small targets in infrared images with complex backgrounds, which is crucial for various applications. However, existing IRSTD methods typically rely solely on image modalities for processing, which fail to fully capture contextual information, leading to limited detection accuracy and adaptability in complex environments. Inspired by vision-language models, this paper proposes a novel framework, SAIST, which integrates textual information with image modalities to enhance IRSTD performance. The framework consists of two main components: Scene Recognition Contrastive Language-Image Pretraining (SR-CLIP) and CLIP-guided Segment Anything Model (CG-SAM). SR-CLIP generates a set of visual descriptions through object-object similarity and object-scene relevance, embedding them into learnable prompts to refine the textual description set. This reduces the domain gap between vision and language, generating precise textual and visual prompts. CG-SAM utilizes the prompts generated by SR-CLIP to accurately guide the Mask Decoder in learning prior knowledge of background features, while incorporating infrared imaging equations to improve small target recognition in complex backgrounds and significantly reduce the false alarm rate. Additionally, this paper introduces the first multimodal IRSTD dataset, MIRSTD, which contains abundant image-text pairs. Experimental results demonstrate that the proposed SAIST method outperforms existing state-of-the-art approaches.

Recent works favored dense signals (e.g., depth, DensePose), as an alternative to sparse signals (e.g., OpenPose), to provide detailed spatial guidance for pose-guided text-to-image generation. However, dense representations raised new challenges, including editing difficulties and potential inconsistencies with textual prompts. This fact motivates us to revisit sparse signals for pose guidance, owing to their simplicity and shape-agnostic nature, which remains underexplored. This paper proposes a novel Spatial-Pose ControlNet(SP-Ctrl), equipping sparse signals with robust controllability for pose-guided image generation. Specifically, we extend OpenPose to a learnable spatial representation, making keypoint embeddings discriminative and expressive. Additionally, we introduce keypoint concept learning, which encourages keypoint tokens to attend to the spatial positions of each keypoint, thus improving pose alignment. Experiments on animal- and human-centric image generation tasks demonstrate that our method outperforms recent spatially controllable T2I generation approaches under sparse-pose guidance and even matches the performance of dense signal-based methods. Moreover, SP-Ctrl shows promising capabilities in diverse and cross-species generation through sparse signals. Codes will be available at https://github.com/DREAMXFAR/SP-Ctrl.

Large-scale multi-modal models have demonstrated remarkable performance across various visual recognition tasks by leveraging extensive paired multi-modal training data. However, in real-world applications, the presence of missing or incomplete modality inputs often leads to significant performance degradation. Recent research has focused on prompt-based strategies to tackle this issue; however, existing methods are hindered by two major limitations: (1) static prompts lack the flexibility to adapt to varying missing-data conditions, and (2) basic prompt-tuning methods struggle to ensure reliable performance when critical modalities are missing.To address these challenges, we propose a novel Synergistic Prompting (SyP) framework for robust visual recognition with missing modalities. The proposed SyP introduces two key innovations: (I) a Dynamic Adapter, which computes adaptive scaling factors to dynamically generate prompts, replacing static parameters for flexible multi-modal adaptation, and (II) a Synergistic Prompting Strategy, which combines static and dynamic prompts to balance information across modalities, ensuring robust reasoning even when key modalities are missing. The proposed SyP achieves significant performance improvements over existing approaches across three widely-used visual recognition datasets, demonstrating robustness under diverse missing rates and conditions. Extensive experiments and ablation studies validate its effectiveness in handling missing modalities, highlighting its superior adaptability and reliability.

Time Series Forecasting (TSF) is critical in many real-world domains like financial planning and health mon-itoring. Recent studies have revealed that Large Language Models (LLMs), with their powerful in-contextual modeling capabilities, hold significant potential for TSF. However, existing LLM-based methods usually perform suboptimally because they neglect the inherent characteristics of time series data. Unlike the textual data used in LLM pre-training, the time series data is semantically sparse and comprises distinctive tempo-ral patterns. To address this problem, we propose LLM-PS to empower the LLM for TSF by learning the fundamental Patterns and meaningful Semantics from time series data. Our LLM-PS incorporates a new multi-scale convolutional neural network adept at capturing both short-term fluctuations and long-term trends within the time series. Meanwhile, we introduce a time-to-text module for extracting valuable semantics across continuous time intervals rather than isolated time points. By integrating these patterns and semantics, LLM - PS effectively models temporal dependencies, enabling a deep comprehension of time series and delivering accurate forecasts. Intensive exper-imental results demonstrate that LLM-PS achieves state-of-the-art performance in both short- and long-term forecasting tasks, as well as in few- and zero-shot settings. Code is available at https://github.com/tangjialiang97ILLMPS.

Robust high-definition (HD) map construction is vital for autonomous driving, yet existing methods often struggle with incomplete multi-view camera data. This paper presents SafeMap, a novel framework specifically designed to ensure accuracy even when certain camera views are missing. SafeMap integrates two key components: the Gaussian-based Perspective View Reconstruction (G-PVR) module and the Distillation-based Bird’s-Eye-View (BEV) Correction (D-BEVC) module. G-PVR leverages prior knowledge of view importance to dynamically prioritize the most informative regions based on the relationships among available camera views. Furthermore, D-BEVC utilizes panoramic BEV features to correct the BEV representations derived from incomplete observations. Together, these components facilitate comprehensive data reconstruction and robust HD map generation. SafeMap is easy to implement and integrates seamlessly into existing systems, offering a plug-and-play solution for enhanced robustness. Experimental results demonstrate that SafeMap significantly outperforms previous methods in both complete and incomplete scenarios, highlighting its superior performance and resilience.

3D visual grounding aims to localize target objects in point clouds based on free-form natural language, which often describes both target and reference objects. Effective alignment between visual and text features is crucial for this task. However, existing two-stage methods that rely solely on object-level features can yield suboptimal accuracy, while one-stage methods that align only point-level features can be prone to noise. In this paper, we propose DDPA-3DVG, a novel framework that progressively aligns visual locations and language descriptions at multiple granularities. Specifically, we decouple natural language descriptions into distinct representations of target objects, reference objects, and their mutual relationships, while disentangling 3D scenes into object-level, voxel-level, and point-level features. By progressively fusing these dual-decoupled features from coarse to fine, our method enhances cross-modal alignment and achieves state-of-the-art performance on three challenging benchmarks—ScanRefer, Nr3D, and Sr3D. The code will be released at https://github.com/HDU-VRLab/DDPA-3DVG.

Massive human-related data is collected to train neural networks for computer vision tasks. A major conflict is exposed relating to software engineers between better developing AI systems and distancing from the sensitive training data. To reconcile this conflict, the paper proposes an efficient privacy-preserving learning paradigm, where images are encrypted to become ``human-imperceptible, machine-recognizable'' via one of the two encryption strategies: (1) random shuffling equally-sized patches and (2) mixing-up sub-patches. Then, minimal adaptations are made to vision transformer to enable it to learn on the encrypted images for vision tasks, including image classification and object detection. Extensive experiments on ImageNet and COCO show that the proposed paradigm achieves comparable accuracy with the competitive methods. Decrypting the encrypted images requires solving an NP-hard jigsaw puzzle or ill-posed inverse problem, which is empirically shown intractable to be recovered by various attackers, including the powerful vision transformer-based attacker. We thus show that the proposed paradigm can ensure the encrypted images have become human-imperceptible while preserving machine-recognizable information.

3D Single Object Tracking (SOT) is a fundamental task in computer vision and plays a critical role in applications like autonomous driving. However, existing algorithms often involve complex designs and multiple loss functions, making model training and deployment challenging. Furthermore, their reliance on fixed probability distribution assumptions (e.g., Laplacian or Gaussian) hinders their ability to adapt to diverse target characteristics such as varying sizes and motion patterns, ultimately affecting tracking precision and robustness. To address these issues, we propose BEVTrack, a simple yet effective motion-based tracking method. BEVTrack directly estimates object motion in Bird's-Eye View (BEV) using a single regression loss. To enhance accuracy for targets with diverse attributes, it learns adaptive likelihood functions tailored to individual targets, avoiding the limitations of fixed distribution assumptions in previous methods. This approach provides valuable priors for tracking and significantly boosts performance. Comprehensive experiments on KITTI, NuScenes, and Waymo Open Dataset demonstrate that BEVTrack achieves state-of-the-art results while operating at 200 FPS, enabling real-time applicability. The code will be released at https://github.com/xmm-prio/BEVTrack.

In this work, we address the new challenge of open-vocabulary fine-grained hand action detection, which aims to recognize hand actions from both known and novel categories using textual descriptions. Traditional hand action detection methods are limited to closed-set detection, making it difficult for them to generalize to new, unseen hand action categories. While current open-vocabulary detection (OVD) methods are effective at detecting novel objects, they face challenges with fine-grained action recognition, particularly when data is limited and heterogeneous. This often leads to poor generalization and performance bias between base and novel categories. To address these issues, we propose a novel approach, Open-FGHA (Open-vocabulary Fine-Grained Hand Action), which learns to distinguish fine-grained features across multiple modalities from limited heterogeneous data. It then identifies optimal matching relationships among these features, enabling accurate open-vocabulary fine-grained hand action detection. Specifically, we introduce three key components: Hierarchical Heterogeneous Low-Rank Adaptation, Bidirectional Selection and Fusion Mechanism, and Cross-Modality Query Generator. These components work in unison to enhance the alignment and fusion of multimodal fine-grained features. Extensive experiments demonstrate that Open-FGHA outperforms existing OVD methods, showing its strong potential for open-vocabulary hand action detection. The source code is available at OV-FGHAD.

Ultra-high-resolution (UHR) remote sensing (RS) imagery offers valuable data for Earth observation but pose challenges for existing multimodal foundation models due to two key bottlenecks: (1) limited availability of UHR training data, and (2) token explosion caused by the large image size. To address data scarcity, we introduce **SuperRS-VQA** (avg. 8,376$\times$8,376) and **HighRS-VQA** (avg. 2,000$\times$1,912), the highest-resolution vision-language datasets in RS to date, covering 22 real-world dialogue tasks. To mitigate token explosion, our pilot studies reveal significant redundancy in RS images: crucial information is concentrated in a small subset of object-centric tokens, while pruning background tokens (e.g., ocean or forest) can even improve performance. Motivated by these findings, we propose two strategies: *Background Token Pruning* and *Anchored Token Selection*, to reduce the memory footprint while preserving key semantics. Integrating these techniques, we introduce **GeoLLaVA-8K**, the first RS-focused multimodal large language model capable of handling inputs up to 8K$\times$8K resolution, built on the LLaVA framework. Trained on SuperRS-VQA and HighRS-VQA, GeoLLaVA-8K sets a new state-of-the-art on the XLRS-Bench. Datasets and code were released at https://github.com/MiliLab/GeoLLaVA-8K.

Recent advances in self-supervised learning for Vision Transformers (ViTs) have fueled breakthroughs in remote sensing (RS) foundation models. However, the quadratic complexity of self-attention poses a significant barrier to scalability, particularly for large models and high-resolution images. While the linear-complexity Mamba architecture offers a promising alternative, existing RS applications of Mamba remain limited to supervised tasks on small, domain-specific datasets. To address these challenges, we propose RoMA, a framework that enables scalable self-supervised pretraining of Mamba-based RS foundation models using large-scale, diverse, unlabeled data. RoMA enhances scalability for high-resolution images through a tailored auto-regressive learning strategy, incorporating two key innovations: 1) a rotation-aware pretraining mechanism combining adaptive cropping with angular embeddings to handle sparsely distributed objects with arbitrary orientations, and 2) multi-scale token prediction objectives that address the extreme variations in object scales inherent to RS imagery. Systematic empirical studies validate that Mamba adheres to RS data and parameter scaling laws, with performance scaling reliably as model and data size increase. Furthermore, experiments across scene classification, object detection, and semantic segmentation tasks demonstrate that RoMA-pretrained Mamba models consistently outperform ViT-based counterparts in both accuracy and computational efficiency. The source code and pretrained models have be released at https://github.com/MiliLab/RoMA.

Diffusion models have achieved remarkable progress in universal image restoration. However, existing methods perform naive inference in the reverse process, which leads to cumulative errors under limited sampling steps and large step intervals. Moreover, they struggle to balance the commonality of degradation representations with restoration quality, often depending on complex compensation mechanisms that enhance fidelity at the expense of efficiency. To address these challenges, we introduce \textbf{DGSolver}, a diffusion generalist solver with universal posterior sampling. We first derive the exact ordinary differential equations for generalist diffusion models to unify degradation representations and design tailored high-order solvers with a queue-based accelerated sampling strategy to improve both accuracy and efficiency. We then integrate universal posterior sampling to better approximate manifold-constrained gradients, yielding a more accurate noise estimation and correcting errors in inverse inference. Extensive experiments demonstrate that DGSolver outperforms state-of-the-art methods in restoration accuracy, stability, and scalability, both qualitatively and quantitatively. Code and models are publicly available at https://github.com/MiliLab/DGSolver.

Current camouflaged object detection methods predominantly follow discriminative segmentation paradigms and heavily rely on predefined categories present in the training data, limiting their generalization to unseen or emerging camouflage objects. This limitation is further compounded by the labor-intensive and time-consuming nature of collecting camouflage imagery. Although Large Vision-Language Models (LVLMs) show potential to improve such issues with their powerful generative capabilities, their understanding of camouflage scenes is still insufficient. To bridge this gap, we introduce MMCSBench, the first comprehensive multimodal benchmark designed to evaluate and advance LVLM capabilities in camouflage scenes. MMCSBench comprises 22,537 images and 76,843 corresponding image-text pairs across five fine-grained camouflage tasks. Additionally, we propose a new task, Camouflage Efficacy Assessment (CEA), aimed at quantitatively evaluating the camouflage effectiveness of objects in images and enabling automated collection of camouflage images from large-scale databases. Extensive experiments on 26 LVLMs reveal significant shortcomings in models' ability to perceive and interpret camouflage scenes. These findings highlight the fundamental differences between natural and camouflaged visual inputs, offering insights for future research in advancing LVLM capabilities within this challenging domain.

Transformer-based sequential recommendation (TSR) models have shown superior performance in recommendation systems, where the quality of item representations plays a crucial role. Classical representation methods integrate item features using concatenation or neural networks to generate homogeneous representation sequences. While straightforward, these methods overlook the heterogeneity of item features, limiting the transformer's ability to capture fine-grained patterns and restricting scalability. Recent studies have attempted to integrate user-side heterogeneous features into item representation sequences, but item-side heterogeneous features, which are vital for performance, remain excluded. To address these challenges, we propose a Heterogeneous Information Transformer model for Sequential Recommendation (HeterRec), which incorporates Heterogeneous Token Flatten Layer (HTFL) and Hierarchical Causal Transformer Layer (HCT). Our HTFL is a novel item tokenization method that converts items into a heterogeneous token set and organizes these tokens into heterogeneous sequences, effectively enhancing performance gains when scaling up the model. Moreover, HCT introduces token-level and item-level causal transformers to extract fine-grained patterns from the heterogeneous sequences. Experiments on offline and online datasets show that the HeterRec model achieves superior performance.

Multi-objective embedding-based retrieval (EBR) has become increasingly critical due to the growing complexity of user behaviors and commercial objectives. While traditional approaches often suffer from data sparsity and limited information sharing between objectives, recent methods utilizing a shared network alongside dedicated sub-networks for each objective partially address these limitations. However, such methods significantly increase the model parameters, leading to an increased retrieval latency and a limited ability to model causal relationships between objectives. To address these challenges, we propose the Cascaded Selective Mask Fine-Tuning (CSMF), a novel method that enhances both retrieval efficiency and serving performance for multi-objective EBR. The CSMF framework selectively masks model parameters to free up independent learning space for each objective, leveraging the cascading relationships between objectives during the sequential fine-tuning. Without increasing network parameters or online retrieval overhead, CSMF computes a linearly weighted fusion score for multiple objective probabilities while supporting flexible adjustment of each objective's weight across various recommendation scenarios. Experimental results on real-world datasets demonstrate the superior performance of CSMF, and online experiments validate its significant practical value.

With the rapid growth of user historical behavior data, user interest modeling has become a prominent aspect in Click-Through Rate (CTR) prediction, focusing on learning user intent representations. However, this complexity poses computational challenges, requiring a balance between model performance and acceptable response times for online services. Traditional methods often utilize filtering techniques. These techniques can lead to the loss of significant information by prioritizing top K items based on item attributes or employing low-precision attention mechanisms. In this study, we introduce LREA, a novel attention mechanism that overcomes the limitations of existing approaches while ensuring computational efficiency. LREA leverages low-rank matrix decomposition to optimize runtime performance and incorporates a specially designed loss function to maintain attention capabilities while preserving information integrity. During the inference phase, matrix absorption and pre-storage strategies are employed to effectively meet runtime constraints. The results of extensive offline and online experiments demonstrate that our method outperforms state-of-the-art approaches.

Abstract Few-shot object detection (FSOD) studies how to detect novel objects with few annotated examples effectively. Recently, it has been demonstrated that decent feature embeddings, including the general feature embeddings that are more invariant to visual changes and the specific feature embeddings that are more discriminative for different object classes, are both important for FSOD. However, current methods lack appropriate mechanisms to sensibly cooperate both types of feature embeddings based on their importance to detecting objects of novel classes, which may result in sub-optimal performance. In this paper, to achieve more effective FSOD, we attempt to explicitly encode both general and specific feature embeddings using learnable tensors and apply a Transformer to help better incorporate them in FSOD according to their relations to the input object features. We thus propose a Transformer-based general and specific embedding learning (T-GSEL) method for FSOD. In T-GSEL, learnable tensors are employed in a three-stage pipeline, encoding feature embeddings in general level, intermediate level, and specific level, respectively. In each stage, we apply a Transformer to first model the relations of the corresponding embedding to input object features and then apply the estimated relations to refine the input features. Meanwhile, we further introduce cross-stage connections between embeddings of different stages to make them complement and cooperate with each other, delivering general, intermediate, and specific feature embeddings stage by stage and utilizing them together for feature refinement in FSOD. In practice, a T-GSEL module is easy to inject. Extensive empirical results further show that our proposed T-GSEL method achieves compelling FSOD performance on both PASCAL VOC and MS COCO datasets compared with other state-of-the-art approaches.

Abstract Vision-and-language navigation (VLN) simulates a visual agent that follows natural-language navigation instructions in real-world scenes. Existing approaches have made enormous progress in navigation in new environments, such as beam search, pre-exploration, and dynamic or hierarchical history encoding. To balance generalization and efficiency, we resort to memorizing visited scenarios apart from the ongoing route while navigating. In this work, we introduce a mechanism of Episodic Scene memory (ESceme) for VLN that wakes an agent’s memories of past visits when it enters the current scene. The episodic scene memory allows the agent to envision a bigger picture of the next prediction. This way, the agent learns to utilize dynamically updated information instead of merely adapting to the current observations. We provide a simple yet effective implementation of ESceme by enhancing the accessible views at each location and progressively completing the memory while navigating. We verify the superiority of ESceme on short-horizon (R2R), long-horizon (R4R), and vision-and-dialog (CVDN) VLN tasks. Our ESceme also wins first place on the CVDN leaderboard. Code is available: https://github.com/qizhust/esceme .

This paper investigates indoor point cloud semantic segmentation under scene-level annotation, which is less explored compared to methods relying on sparse point-level labels. In the absence of precise point-level labels, current methods first generate point-level pseudo-labels, which are then used to train segmentation models. However, generating accurate pseudo-labels for each point solely based on scene-level annotations poses a considerable challenge, substantially affecting segmentation performance. Consequently, to enhance accuracy, this paper proposes a high-quality pseudo-label generation framework by exploring contemporary multi-modal information and region-point semantic consistency. Specifically, with a cross-modal feature guidance module, our method utilizes 2D-3D correspondences to align point cloud features with corresponding 2D image pixels, thereby assisting point cloud feature learning. To further alleviate the challenge presented by the scene-level annotation, we introduce a region-point semantic consistency module. It produces regional semantics through a region-voting strategy derived from point-level semantics, which are subsequently employed to guide the point-level semantic predictions. Leveraging the aforementioned modules, our method can rectify inaccurate point-level semantic predictions during training and obtain high-quality pseudo-labels. Significant improvements over previous works on ScanNet v2 and S3DIS datasets under scene-level annotation can demonstrate the effectiveness. Additionally, comprehensive ablation studies validate the contributions of our approach's individual components.

Accurate hyperspectral image (HSI) interpretation is critical for providing valuable insights into various earth observation-related applications such as urban planning, precision agriculture, and environmental monitoring. However, existing HSI processing methods are predominantly task-specific and scene-dependent, which severely limits their ability to transfer knowledge across tasks and scenes, thereby reducing the practicality in real-world applications. To address these challenges, we present HyperSIGMA, a vision transformer-based foundation model that unifies HSI interpretation across tasks and scenes, scalable to over one billion parameters. To overcome the spectral and spatial redundancy inherent in HSIs, we introduce a novel sparse sampling attention (SSA) mechanism, which effectively promotes the learning of diverse contextual features and serves as the basic block of HyperSIGMA. HyperSIGMA integrates spatial and spectral features using a specially designed spectral enhancement module. In addition, we construct a large-scale hyperspectral dataset, HyperGlobal-450K, for pre-training, which contains about 450 K hyperspectral images, significantly surpassing existing datasets in scale. Extensive experiments on various high-level and low-level HSI tasks demonstrate HyperSIGMA's versatility and superior representational capability compared to current state-of-the-art methods. Moreover, HyperSIGMA shows significant advantages in scalability, robustness, cross-modal transferring capability, real-world applicability, and computational efficiency.

The Segment Anything Model (SAM), a profound vision foundation model pretrained on a large-scale dataset, breaks the boundaries of general segmentation and sparks various downstream applications. This paper introduces Hi-SAM, a unified model leveraging SAM for hierarchical text segmentation. Hi-SAM excels in segmentation across four hierarchies, including pixel-level text, word, text-line, and paragraph, while realizing layout analysis as well. Specifically, we first turn SAM into a high-quality pixel-level text segmentation (TS) model through a parameter-efficient fine-tuning approach. We use this TS model to iteratively generate the pixel-level text labels in a semi-automatical manner, unifying labels across the four text hierarchies in the HierText dataset. Subsequently, with these complete labels, we launch the end-to-end trainable Hi-SAM based on the TS architecture with a customized hierarchical mask decoder. During inference, Hi-SAM offers both automatic mask generation (AMG) mode and promptable segmentation (PS) mode. In the AMG mode, Hi-SAM segments pixel-level text foreground masks initially, then samples foreground points for hierarchical text mask generation and achieves layout analysis in passing. As for the PS mode, Hi-SAM provides word, text-line, and paragraph masks with a single point click. Experimental results show the state-of-the-art performance of our TS model: 84.86% fgIOU on Total-Text and 88.96% fgIOU on TextSeg for pixel-level text segmentation. Moreover, compared to the previous specialist for joint hierarchical detection and layout analysis on HierText, Hi-SAM achieves significant improvements: 4.73% PQ and 5.39% F1 on the text-line level, 5.49% PQ and 7.39% F1 on the paragraph level layout analysis, requiring fewer training epochs.

Unsupervised Domain Adaptive Semantic Segmentation (UDA-SS) aims to transfer the supervision from a labeled source domain to an unlabeled and shifted target domain. The majority of existing UDA-SS works typically consider images whilst recent attempts have extended further to tackle videos by modeling the temporal dimension. Although two lines of research share the major challenges - overcoming the underlying domain distribution shift, their studies are largely independent. It causes several issues: (1) The insights gained from each line of research remain fragmented, leading to a lack of holistic understanding of the problem and potential solutions. (2) Preventing the unification of methods and best practices across two scenarios (images and videos) will lead to redundant efforts and missed opportunities for cross-pollination of ideas. (3) Without a unified approach, the knowledge and advancements made in one scenario may not be effectively transferred to the other, leading to suboptimal performance and slower progress. Under this observation, we advocate unifying the study of UDA-SS across video and image scenarios, enabling a more comprehensive understanding, synergistic advancements, and efficient knowledge sharing. To that end, we explore the unified UDA-SS from a general domain augmentation perspective, serving as a unifying framework, enabling improved generalization, and potential for cross-pollination, ultimately contributing to the practical impact and overall progress. Specifically, we propose a Quad-directional Mixup (QuadMix) method, characterized by tackling intra-domain discontinuity, fragmented gap bridging, and feature inconsistencies through four-directional paths designed for intra- and inter-domain mixing within an explicit feature space. To deal with temporal shifts within videos, we incorporate optical flow-guided feature aggregation across spatial and temporal dimensions for fine-grained domain alignment, which is extendable to image scenarios. Extensive experiments show that QuadMix outperforms the state-of-the-art works by large margins on four challenging UDA-SS benchmarks.

RGB-D Salient Object Detection (SOD) aims to segment the most prominent areas and objects in a given pair of RGB and depth images. Most current models adopt a dual-stream structure to extract information from both RGB and depth images. However, this leads to an exponential increase in the number of parameters and computations in the model. Moreover, the discrepancy between RGB pretrained and the 3D geometric relationships in depth maps present a challenge for the encoder in capturing spatial structural details. These issues impact the model's accuracy in locating salient objects and distinguishing edge details. To address these, we propose a novel early feature fusion network, named FasterSal, which enables more efficient RGB-D SOD. FasterSal uses a single stream structure to receive RGB images and depth maps, extracting features based on the 3D geometric relationships in the depth map while fully leveraging the pretrained RGB encoder. This approach effectively avoids the inconsistencies between depth modality and the RGB pretrained encoder. It also significantly reduces the number of network parameters while maintaining efficient feature encoding capabilities. To achieve finer edge learning, the detail-aware loss and texture enhancement module are introduced. These modules are designed to extract latent details in high-frequency component features and to enhance the edge learning capability of the model using distance information. Experimental results on several benchmark datasets confirm the effectiveness and superiority of our method over the state-of-the-art approaches, achieving a good balance between performance and speed with only 3.4 million parameters and a CPU operating speed of 63 FPS.

Infrared small target detection (IRSTD) aims to identify and locate small targets amidst background noise. It is highly valuable in various practical application domains, such as maritime rescue and early warning systems deployed in challenging conditions such as harsh weather, low illumination, and long imaging distances. Different from existing works that either adopt well-designed backbone networks or devise specific modules to improve them from different aspects, in this article, we formulate the learning process of IRSTD from a novel perspective, i.e., the mechanism of pixel movement. Considering that the movement of pixels passing through the layers of the network for IRSTD can be analogized to the flow of particles in a fluid dynamic system, we propose a computational fluid dynamic network (CFD-Net) derived from computational fluid dynamics. Technically, we leverage the superiority of the unilateral difference equation with third-order accuracy and devise a unilateral differential residual structure as the backbone of CFD-Net. This design ensures that the pixel stream only flows in the forward direction. In addition, a switch-controlled multidirectional treatment tank (SMTT) is introduced to CFD-Net to dynamically guide the pixel stream to the appropriate path for different targets with varying shapes and orientations, facilitating learning robust target representation and improving detection performance. The proposed CFD-Net is evaluated on the IRSTD-1k and SIRST datasets and is found to outperform existing state-of-the-art (SOTA) methods.

Composed image retrieval is a type of image retrieval task where the user provides a reference image as a starting point and specifies a text on how to shift from the starting point to the desired target image. However, most existing methods focus on the composition learning of text and reference images and oversimplify the text as a description, neglecting the inherent structure and the user's shifting intention of the texts. As a result, these methods typically take shortcuts that disregard the visual cue of the reference images. To address this issue, we reconsider the text as instructions and propose a Semantic Shift Network (SSN) that explicitly decomposes the semantic shifts into two steps: from the reference image to the visual prototype and from the visual prototype to the target image. Specifically, SSN explicitly decomposes the instructions into two components: degradation and upgradation, where the degradation is used to picture the visual prototype from the reference image, while the upgradation is used to enrich the visual prototype into the final representations to retrieve the desired target image. The experimental results show that the proposed SSN demonstrates a significant improvement of 5.42% and 1.37% on the CIRR and FashionIQ datasets, respectively, and establishes a new state-of-the-art performance. The code is available at https://github.com/starxing-yuu/SSN.

The Segment Anything Model (SAM) is a powerful foundation model that has revolutionised image segmentation. To apply SAM to surgical instrument segmentation, a common approach is to locate precise points or boxes of instruments and then use them as prompts for SAM in a zero-shot manner. However, we observe two problems with this naive pipeline: (1) the domain gap between natural objects and surgical instruments leads to inferior generalisation of SAM; and (2) SAM relies on precise point or box locations for accurate segmentation, requiring either extensive manual guidance or a well-performing specialist detector for prompt preparation, which leads to a complex multi-stage pipeline. To address these problems, we introduce SurgicalSAM, a novel end-to-end efficient-tuning approach for SAM to effectively integrate surgical-specific information with SAM’s pre-trained knowledge for improved generalisation. Specifically, we propose a lightweight prototype-based class prompt encoder for tuning, which directly generates prompt embeddings from class prototypes and eliminates the use of explicit prompts for improved robustness and a simpler pipeline. In addition, to address the low inter-class variance among surgical instrument categories, we propose contrastive prototype learning, further enhancing the discrimination of the class prototypes for more accurate class prompting. The results of extensive experiments on both EndoVis2018 and EndoVis2017 datasets demonstrate that SurgicalSAM achieves state-of-the-art performance while only requiring a small number of tunable parameters. The source code is available at https://github.com/wenxi-yue/SurgicalSAM.

Infrared Small Target Detection (IRSTD) refers to detecting faint targets in infrared images, which has achieved notable progress with the advent of deep learning. However, the drive for improved detection accuracy has led to larger, intricate models with redundant parameters, causing storage and computation inefficiencies. In this pioneering study, we introduce the concept of utilizing network pruning to enhance the efficiency of IRSTD. Due to the challenge posed by low signal-to-noise ratios and the absence of detailed semantic information in infrared images, directly applying existing pruning techniques yields suboptimal performance. To address this, we propose a novel wavelet structure-regularized soft channel pruning method, giving rise to the efficient IRPruneDet model. Our approach involves representing the weight matrix in the wavelet domain and formulating a wavelet channel pruning strategy. We incorporate wavelet regularization to induce structural sparsity without incurring extra memory usage. Moreover, we design a soft channel reconstruction method that preserves important target information against premature pruning, thereby ensuring an optimal sparse structure while maintaining overall sparsity. Through extensive experiments on two widely-used benchmarks, our IRPruneDet method surpasses established techniques in both model complexity and accuracy. Specifically, when employing U-net as the baseline network, IRPruneDet achieves a 64.13% reduction in parameters and a 51.19% decrease in FLOPS, while improving IoU from 73.31% to 75.12% and nIoU from 70.92% to 74.30%. The code is available at https://github.com/hd0013/IRPruneDet.

Multi-view camera-based 3D object detection has become popular due to its low cost, but accurately inferring 3D geometry solely from camera data remains challenging and may lead to inferior performance. Although distilling precise 3D geometry knowledge from LiDAR data could help tackle this challenge, the benefits of LiDAR information could be greatly hindered by the significant modality gap between different sensory modalities. To address this issue, we propose a Simulated multi-modal Distillation (SimDistill) method by carefully crafting the model architecture and distillation strategy. Specifically, we devise multi-modal architectures for both teacher and student models, including a LiDAR-camera fusion-based teacher and a simulated fusion-based student. Owing to the ``identical'' architecture design, the student can mimic the teacher to generate multi-modal features with merely multi-view images as input, where a geometry compensation module is introduced to bridge the modality gap. Furthermore, we propose a comprehensive multi-modal distillation scheme that supports intra-modal, cross-modal, and multi-modal fusion distillation simultaneously in the Bird's-eye-view space. Incorporating them together, our SimDistill can learn better feature representations for 3D object detection while maintaining a cost-effective camera-only deployment. Extensive experiments validate the effectiveness and superiority of SimDistill over state-of-the-art methods, achieving an improvement of 4.8% mAP and 4.1% NDS over the baseline detector. The source code will be released at https://github.com/ViTAE-Transformer/SimDistill.

Existing research based on deep learning has extensively explored the problem of daytime image dehazing. However, few studies have considered the characteristics of nighttime hazy scenes. There are two distinctions between nighttime and daytime haze. First, there may be multiple active col-ored light sources with lower illumination intensity in night-time scenes, which may cause haze, glow and noise with localized, coupled and frequency inconsistent characteris-tics. Second, due to the domain discrepancy between simulated and real-world data, unrealistic brightness may occur when applying a dehazing model trained on simulated data to real-world data. To address the above two issues, we propose a semi-supervised model for real-world nighttime dehazing. First, the spatial attention and frequency spectrum filtering are implemented as a spatial-frequency do-main information interaction module to handle the first is-sue. Second, a pseudo-label-based retraining strategy and a local window-based brightness loss for semi-supervised training process is designed to suppress haze and glow while achieving realistic brightness. Experiments on public benchmarks validate the effectiveness of the proposed method and its superiority over state-of-the-art methods. The source code and Supplementary Materials are placed in the https://github.com/Xiaofeng-life/SFSNiD.

LiDAR semantic segmentation (LSS) is a critical task in autonomous driving and has achieved promising progress. However, prior LSS methods are conventionally investigated and evaluated on datasets within the same domain in clear weather. The robustness of LSS models in unseen scenes and all weather conditions is crucial for ensuring safety and reliability in real applications. To this end, we propose UniMix, a universal method that enhances the adaptability and generalizability of LSS models. UniMix first leverages physically valid adverse weather simulation to construct a Bridge Domain, which serves to bridge the domain gap between the clear weather scenes and the adverse weather scenes. Then, a Universal Mixing operator is defined regarding spatial, intensity, and semantic distributions to create the intermediate domain with mixed samples from given domains. Integrating the proposed two techniques into a teacher-student framework, UniMix efficiently mitigates the domain gap and enables LSS models to learn weather-robust and domain-invariant representations. We devote UniMix to two main setups: 1) unsupervised domain adaption, adapting the model from the clear weather source domain to the adverse weather target domain; 2) domain generalization, learning a model that generalizes well to unseen scenes in adverse weather. Extensive experiments validate the effectiveness of UniMix across different tasks and datasets, all achieving superior performance over state-of-the-art methods. The code will be released.

Diffusion models have achieved remarkable success in generating realistic images but suffer from generating accurate human hands, such as incorrect finger counts or irregular shapes. This difficulty arises from the complex task of learning the physical structure and pose of hands from training images, which involves extensive deformations and occlusions. For correct hand generation, our paper introduces a lightweight post-processing solution called HandRefiner. HandRefiner employs a conditional inpainting approach to rectify malformed hands while leaving other parts of the image untouched. We leverage the hand mesh reconstruction model that consistently adheres to the correct number of fingers and hand shape, while also being capable of fitting the desired hand pose in the generated image. Given a generated failed image due to malformed hands, we utilize ControlNet modules to re-inject such correct hand information. Additionally, we uncover a phase transition phenomenon within ControlNet as we vary the control strength. It enables us to take advantage of more readily available synthetic data without suffering from the domain gap between realistic and synthetic hands. Experiments demonstrate that HandRefiner can significantly improve the generation quality quantitatively and qualitatively. The code is available at https://github.com/wenquanlu/HandRefiner.

Neural implicit representations have recently revolutionized simultaneous localization and mapping (SLAM), giving rise to a groundbreaking paradigm known as NeRF-based SLAM. However, existing methods often fall short in accurately estimating poses and reconstructing scenes. This limitation largely stems from their reliance on volume rendering techniques, which oversimplify the modeling process. In this paper, we introduce a novel neural implicit SLAM system named SAR-SLAM to address these shortcomings. Our approach reconstructs Neural Radiance Fields (NeRFs) using a self-attentive architecture and represents scenes through neural point cloud encoding. Unlike previous NeRF-based SLAM methods, which depend on traditional volume rendering equations for scene representation and view synthesis, our method employs a self-attentive rendering framework with the Transformer architecture during mapping and tracking stages. To enable incremental mapping, we anchor scene features within a neural point cloud, striking a balance between estimation accuracy and computational cost. Experimental results on three challenging datasets show the superior performance and robustness of our SAR-SLAM compared to recent NeRF-based SLAM systems. The code will be released.

Recent advancements in deep learning have greatly advanced the field of infrared small object detection (IRSTD). Despite their remarkable success, a notable gap persists between these IRSTD methods and generic segmentation approaches in natural image domains. This gap primarily arises from the significant modality differences and the limited availability of infrared data. In this study, we aim to bridge this divergence by investigating the adaptation of generic segmentation models, such as the Segment Anything Model (SAM), to IRSTD tasks. Our investigation reveals that many generic segmentation models can achieve comparable performance to state-of-the-art IRSTD methods. However, their full potential in IRSTD remains untapped. To address this, we propose a simple, lightweight, yet effective baseline model for segmenting small infrared objects. Through appropriate distillation strategies, we empower smaller student models to outperform state-of-the-art methods, even surpassing fine-tuned teacher results. Furthermore, we enhance the model's performance by introducing a novel query design comprising dense and sparse queries to effectively encode multi-scale features. Through extensive experimentation across four popular IRSTD datasets, our model demonstrates significantly improved performance in both accuracy and throughput compared to existing approaches, surpassing SAM and Semantic-SAM by over 14 IoU on NUDT and 4 IoU on IRSTD1k. The source code and models will be released at https://github.com/O937-blip/SimIR.

Detecting hand actions in videos is crucial for understanding video content and has diverse real-world applications. Existing approaches often focus on whole-body actions or coarse-grained action categories, lacking fine-grained hand-action localization information. To fill this gap, we introduce the FHA-Kitchens (Fine-Grained Hand Actions in Kitchen Scenes) dataset, providing both coarse- and fine-grained hand action categories along with localization annotations. This dataset comprises 2,377 video clips and 30,047 frames, annotated with approximately 200k bounding boxes and 880 action categories. Evaluation of existing action detection methods on FHA-Kitchens reveals varying generalization capabilities across different granularities. To handle multi-granularity in hand actions, we propose MG-HAD, an End-to-End Multi-Granularity Hand Action Detection method. It incorporates two new designs: Multi-dimensional Action Queries and Coarse-Fine Contrastive Denoising. Extensive experiments demonstrate MG-HAD's effectiveness for multi-granularity hand action detection, highlighting the significance of FHA-Kitchens for future research and real-world applications. The dataset and source code are available at MG-HAD.

Driving systems often rely on high-definition (HD) maps for precise environmental information, which is crucial for planning and navigation. While current HD map constructors perform well under ideal conditions, their resilience to real-world challenges, \eg, adverse weather and sensor failures, is not well understood, raising safety concerns. This work introduces MapBench, the first comprehensive benchmark designed to evaluate the robustness of HD map construction methods against various sensor corruptions. Our benchmark encompasses a total of 29 types of corruptions that occur from cameras and LiDAR sensors. Extensive evaluations across 31 HD map constructors reveal significant performance degradation of existing methods under adverse weather conditions and sensor failures, underscoring critical safety concerns. We identify effective strategies for enhancing robustness, including innovative approaches that leverage multi-modal fusion, advanced data augmentation, and architectural techniques. These insights provide a pathway for developing more reliable HD map construction methods, which are essential for the advancement of autonomous driving technology. The benchmark toolkit and affiliated code and model checkpoints have been made publicly accessible.

Beyond the text detection and recognition tasks in image text spotting, video text spotting presents an augmented challenge with the inclusion of tracking. While advanced end-to-end trainable methods have shown commendable performance, the pursuit of multi-task optimization may pose the risk of producing sub-optimal outcomes for individual tasks. In this paper, we identify a main bottleneck in the state-of-the-art video text spotter: the limited recognition capability. In response to this issue, we propose to efficiently turn an off-the-shelf query-based image text spotter into a specialist on video and present a simple baseline termed GoMatching, which focuses the training efforts on tracking while maintaining strong recognition performance. To adapt the image text spotter to video datasets, we add a rescoring head to rescore each detected instance's confidence via efficient tuning, leading to a better tracking candidate pool. Additionally, we design a long-short term matching module, termed LST-Matcher, to enhance the spotter's tracking capability by integrating both long- and short-term matching results via Transformer. Based on the above simple designs, GoMatching delivers new records on ICDAR15-video, DSText, BOVText, and our proposed novel test set with arbitrary-shaped text termed ArTVideo, which demonstates GoMatching's capability to accommodate general, dense, small, arbitrary-shaped, Chinese and English text scenarios while saving considerable training budgets. The code will be released.

Abstract Deep models trained by using clean data have achieved tremendous success in fine-grained image classification. Yet, they generally suffer from significant performance degradation when encountering noisy labels. Existing approaches to handle label noise, though proved to be effective for generic object recognition, usually fail on fine-grained data. The reason is that, on fine-grained data, the category difference is subtle and the training sample size is small. Then deep models could easily overfit the noisy labels. To improve the robustness of deep models on noisy data for fine-grained visual categorization, in this paper, we propose a novel learning framework named ProtoSimi. Our method employs an adaptive label correction strategy, ensuring effective learning on limited data. Specifically, our approach considers the criteria of exploring the effectiveness of both global class-prototype and part class-prototype similarities in identifying and correcting labels of samples. We evaluate our method on three standard benchmarks of fine-grained recognition. Experimental results show that our method outperforms the existing label noisy methods by a large margin. In ablation studies, we also verify that our method is non-sensitive to hyper-parameters selection and can be integrated with other FGVC methods to increase the generalization performance.

Deep supervised learning algorithms typically require a large volume of labeled data to achieve satisfactory performance. However, the process of collecting and labeling such data can be expensive and time-consuming. Self-supervised learning (SSL), a subset of unsupervised learning, aims to learn discriminative features from unlabeled data without relying on human-annotated labels. SSL has garnered significant attention recently, leading to the development of numerous related algorithms. However, there is a dearth of comprehensive studies that elucidate the connections and evolution of different SSL variants. This paper presents a review of diverse SSL methods, encompassing algorithmic aspects, application domains, three key trends, and open research questions. First, we provide a detailed introduction to the motivations behind most SSL algorithms and compare their commonalities and differences. Second, we explore representative applications of SSL in domains such as image processing, computer vision, and natural language processing. Lastly, we discuss the three primary trends observed in SSL research and highlight the open questions that remain.

Vision Transformers (ViTs) have achieved impressive performance over various computer vision tasks. However, modeling global correlations with multi-head self-attention (MSA) layers leads to two widely recognized issues: the massive computational resource consumption and the lack of intrinsic inductive bias for modeling local visual patterns. To solve both issues, we devise a simple yet effective method named Single-Path Vision Transformer pruning (SPViT), to efficiently and automatically compress the pre-trained ViTs into compact models with proper locality added. Specifically, we first propose a novel weight-sharing scheme between MSA and convolutional operations, delivering a single-path space to encode all candidate operations. In this way, we cast the operation search problem as finding which subset of parameters to use in each MSA layer, which significantly reduces the computational cost and optimization difficulty, and the convolution kernels can be well initialized using pre-trained MSA parameters. Relying on the single-path space, we introduce learnable binary gates to encode the operation choices in MSA layers. Similarly, we further employ learnable gates to encode the fine-grained MLP expansion ratios of FFN layers. In this way, our SPViT optimizes the learnable gates to automatically explore from a vast and unified search space and flexibly adjust the MSA-FFN pruning proportions for each individual dense model. We conduct extensive experiments on two representative ViTs showing that our SPViT achieves a new SOTA for pruning on ImageNet-1 k. For example, our SPViT can trim 52.0% FLOPs for DeiT-B and get an impressive 0.6% top-1 accuracy gain simultaneously.

In this paper, we show the surprisingly good properties of plain vision transformers for body pose estimation from various aspects, namely simplicity in model structure, scalability in model size, flexibility in training paradigm, and transferability of knowledge between models, through a simple baseline model dubbed ViTPose. ViTPose employs the plain and non-hierarchical vision transformer as an encoder to encode features and a lightweight decoder to decode body keypoints in either a top-down or a bottom-up manner. It can be scaled to 1B parameters by taking the advantage of the scalable model capacity and high parallelism, setting a new Pareto front for throughput and performance. Besides, ViTPose is very flexible regarding the attention type, input resolution, and pre-training and fine-tuning strategy. Based on the flexibility, a novel ViTPose++ model is proposed to deal with heterogeneous body keypoint categories via knowledge factorization, i.e., adopting task-agnostic and task-specific feed-forward networks in the transformer. We also demonstrate that the knowledge of large ViTPose models can be easily transferred to small ones via a simple knowledge token. Our largest single model ViTPose-G sets a new record on the MS COCO test set without model ensemble. Furthermore, our ViTPose++ model achieves state-of-the-art performance simultaneously on a series of body pose estimation tasks, including MS COCO, AI Challenger, OCHuman, MPII for human keypoint detection, COCO-Wholebody for whole-body keypoint detection, as well as AP-10K and APT-36K for animal keypoint detection, without sacrificing inference speed.

Texture recognition is a challenging visual task since its multiple primitives or attributes can be perceived from the texture image under different spatial contexts. Existing approaches predominantly built upon CNN incorporate rich local descriptors with orderless aggregation to capture invariance to the spatial layout. However, these methods ignore the inherent structure relation organized by primitives and the semantic concept described by attributes, which are critical cues for texture representation. In this paper, we propose a novel Multiple Primitives and Attributes Perception network (MPAP) that extracts features by modeling the relation of bottom-up structure and top-down attribute in a multi-branch unified framework. A bottom-up process is first proposed to capture the inherent relation of various primitive structures by leveraging structure dependency and spatial order information. Then, a top-down process is introduced to model the latent relation of multiple attributes by transferring attribute-related features between adjacent branches. Moreover, an augmentation module is devised to bridge the gap between high-level attributes and low-level structure features. MPAP can learn representation through jointing bottom-up and top-down processes in a mutually reinforced manner. Experimental results on six challenging texture datasets demonstrate the superiority of MPAP over state-of-the-art methods in terms of accuracy, robustness, and efficiency.

Window-based attention has become a popular choice in vision transformers due to its superior performance, lower computational complexity, and less memory footprint. However, the design of hand-crafted windows, which is data-agnostic, constrains the flexibility of transformers to adapt to objects of varying sizes, shapes, and orientations. To address this issue, we propose a novel quadrangle attention (QA) method that extends the window-based attention to a general quadrangle formulation. Our method employs an end-to-end learnable quadrangle regression module that predicts a transformation matrix to transform default windows into target quadrangles for token sampling and attention calculation, enabling the network to model various targets with different shapes and orientations and capture rich context information. We integrate QA into plain and hierarchical vision transformers to create a new architecture named QFormer, which offers minor code modifications and negligible extra computational cost. Extensive experiments on public benchmarks demonstrate that QFormer outperforms existing representative vision transformers on various vision tasks, including classification, object detection, semantic segmentation, and pose estimation. The code will be made publicly available at QFormer.

Emotions can be perceived from a person's gait, i.e., their walking style. Existing methods on gait emotion recognition mainly leverage the posture information as input, but ignore the body movement, which contains complementary information for recognizing emotions evoked in the gait. In this paper, we propose a Bilateral Posture and Movement Graph Convolutional Network (BPM-GCN) that consists of two parallel streams, namely posture stream and movement stream, to recognize emotions from two views. The posture stream aims to explicitly analyse the emotional state of the person. Specifically, we design a novel regression constraint based on the hand-engineered features to distill the prior affective knowledge into the network and boost the representation learning. The movement stream is designed to describe the intensity of the emotion, which is an implicitly cue for recognizing emotions. To achieve this goal, we employ a higher-order velocity-acceleration pair to construct graphs, in which the informative movement features are utilized. Besides, we design a PM-Interacted feature fusion mechanism to adaptively integrate the features from the two streams. Therefore, the two streams collaboratively contribute to the performance from two complementary views. Extensive experiments on the largest benchmark dataset Emotion-Gait show that BPM-GCN performs favorably against the state-of-the-art approaches (with at least$4.59\%$performance improvement). The source code is released onhttps://github.com/exped1230/BPM-GCN.

Reconstructing a high-resolution hyperspectral image (HSI) from a low-resolution HSI is significant for many applications, such as remote sensing and aerospace. Most deep learning-based HSI super-resolution methods pay more attention to developing novel network structures but rarely study the HSI super-resolution problem from the perspective of image dynamic evolution. In this article, we propose that the HSI pixel motion during the super-resolution reconstruction process can be analogized to the particle movement in the smoothed particle hydrodynamics (SPH) field. To this end, we design an SPH network (SPH-Net) for HSI super-resolution in light of the SPH theory. Specifically, we construct a smooth function based on SPH and design a smooth convolution in multiscales to exploit spectral correlation and preserve the spectral information in the super-resolved image. In addition, we apply the SPH approximation method to discretize the Navier-Stokes motion equation into SPH equation form, which can guide the HSI pixel motion in the desired direction during super-resolution reconstruction, thereby producing clear edges in the spatial domain. Experiments on three public hyperspectral datasets demonstrate that the proposed SPH-Net outperforms the state-of-the-art methods in terms of objective metrics and visual quality.

Recent object re-identification (Re-ID) methods gain high efficiency via lightweight student models trained by knowledge distillation (KD). However, the huge architectural difference between lightweight students and heavy teachers causes students to have difficulties in receiving and understanding teachers' knowledge, thus losing certain accuracy. To this end, we propose a refiner-expander-refiner (RER) structure to enlarge a student's representational capacity and prune the student's complexity. The expander is a multi-branch convolutional layer to expand the student's representational capacity to understand a teacher's knowledge comprehensively, which does not require any feature-dimensional adapter to avoid knowledge distortions. The two refiners are 1×1 convolutional layers to prune the input and output channels of the expander. In addition, in order to alleviate the competition accuracy-related and pruning-related gradients, we design a common consensus gradient resetting (CCGR) method, which discards unimportant channels according to the intersection of each sample's unimportant channel judgment. Finally, the trained RER can be simplified into a slim convolutional layer via re-parameterization to speed up inference. As a result, we propose an expanding and refining hybrid compressing (ERHC) method. Extensive experiments show that our ERHC has superior inference speed and accuracy, e.g., on the VeRi-776 dataset, given the ResNet101 as a teacher, ERHC saves 75.33% model parameters (MP) and 74.29% floating-point of operations (FLOPs) without sacrificing accuracy.

Multimodal aspect-level sentiment analysis has attracted increasing attention in recent years. However, existing methods have two unaddressed limitations: (1) due to the lack of labelled pre-training data of dedicated sentiment analysis, the methods with a pre-training manner produce suboptimal prediction results; (2) most existing methods employ a self-attention encoder to fuse multimodal tokens, which not only ignores the alignment relationship between different modal tokens but also makes the model unable to capture the semantic links between images and texts. In this paper, we propose a momentum contrastive learning network (MOCOLNet) to overcome above limitations. First, we merge the pre-training stage with the training stage to design an end-to-end training manner which uses less labelled data dedicated to sentiment analysis to obtain better prediction results. Second, we propose a multimodal contrastive learning method to align the different modal representations before data fusing, and design a cross-modal matching strategy to provide semantic interactive information between texts and images. Moreover, we introduce an auxiliary momentum strategy to increase the robustness of model. We also analyse the effectiveness of the proposed multimodal contrastive learning method using a mutual information theory. Experiments verify that the proposed MOCOLNet is superior to other strong baselines.

Visual affordance grounding aims to segment all possible interaction regions between people and objects from an image/video, which benefits many applications, such as robot grasping and action recognition. Prevailing methods predominantly depend on the appearance feature of the objects to segment each region of the image, which encounters the following two problems: 1) there are multiple possible regions in an object that people interact with and 2) there are multiple possible human interactions in the same object region. To address these problems, we propose a hand-aided affordance grounding network (HAG-Net) that leverages the aided clues provided by the position and action of the hand in demonstration videos to eliminate the multiple possibilities and better locate the interaction regions in the object. Specifically, HAG-Net adopts a dual-branch structure to process the demonstration video and object image data. For the video branch, we introduce hand-aided attention to enhance the region around the hand in each video frame and then use the long short-term memory (LSTM) network to aggregate the action features. For the object branch, we introduce a semantic enhancement module (SEM) to make the network focus on different parts of the object according to the action classes and utilize a distillation loss to align the output features of the object branch with that of the video branch and transfer the knowledge in the video branch to the object branch. Quantitative and qualitative evaluations on two challenging datasets show that our method has achieved state-of-the-art results for affordance grounding. The source code is available at: https://github.com/lhc1224/HAG-Net.

Recently, the joint learning framework (JOINT) integrates matching based transductive reasoning and online inductive learning to achieve accurate and robust semi-supervised video object segmentation (SVOS). However, using the mask embedding as the label to guide the generation of target features in the two branches may result in inadequate target representation and degrade the performance. Besides, how to reasonably fuse the target features in the two different branches rather than simply adding them together to avoid the adverse effect of one dominant branch has not been investigated. In this paper, we propose a novel framework that emphasizes Learning to Learn Better (LLB) target features for SVOS, termed LLB, where we design the discriminative label generation module (DLGM) and the adaptive fusion module to address these issues. Technically, the DLGM takes the background-filtered frame instead of the target mask as input and adopts a lightweight encoder to generate the target features, which serves as the label of the online few-shot learner and the value of the decoder in the transformer to guide the two branches to learn more discriminative target representation. The adaptive fusion module maintains a learnable gate for each branch, which reweighs the element-wise feature representation and allows an adaptive amount of target information in each branch flowing to the fused target feature, thus preventing one branch from being dominant and making the target feature more robust to distractor. Extensive experiments on public benchmarks show that our proposed LLB method achieves state-of-the-art performance.

Current 3D single object tracking methods are typically based on VoteNet, a 3D region proposal network. Despite the success, using a single seed point feature as the cue for offset learning in VoteNet prevents high-quality 3D proposals from being generated. Moreover, seed points with different importance are treated equally in the voting process, aggravating this defect. To address these issues, we propose a novel global-local transformer voting scheme to provide more informative cues and guide the model pay more attention on potential seed points, promoting the generation of high-quality 3D proposals. Technically, a global-local transformer (GLT) module is employed to integrate object- and patch-aware prior into seed point features to effectively form strong feature representation for geometric positions of the seed points, thus providing more robust and accurate cues for offset learning. Subsequently, a simple yet effective training strategy is designed to train the GLT module. We develop an importance prediction branch to learn the potential importance of the seed points and treat the output weights vector as a training constraint term. By incorporating the above components together, we exhibit a superior tracking method GLT-T. Extensive experiments on challenging KITTI and NuScenes benchmarks demonstrate that GLT-T achieves state-of-the-art performance in the 3D single object tracking task. Besides, further ablation studies show the advantages of the proposed global-local transformer voting scheme over the original VoteNet. Code and models will be available at https://github.com/haooozi/GLT-T.

Recently, Transformer-based methods, which predict polygon points or Bezier curve control points for localizing texts, are popular in scene text detection. However, these methods built upon detection transformer framework might achieve sub-optimal training efficiency and performance due to coarse positional query modeling. In addition, the point label form exploited in previous works implies the reading order of humans, which impedes the detection robustness from our observation. To address these challenges, this paper proposes a concise Dynamic Point Text DEtection TRansformer network, termed DPText-DETR. In detail, DPText-DETR directly leverages explicit point coordinates to generate position queries and dynamically updates them in a progressive way. Moreover, to improve the spatial inductive bias of non-local self-attention in Transformer, we present an Enhanced Factorized Self-Attention module which provides point queries within each instance with circular shape guidance. Furthermore, we design a simple yet effective positional label form to tackle the side effect of the previous form. To further evaluate the impact of different label forms on the detection robustness in real-world scenario, we establish an Inverse-Text test set containing 500 manually labeled images. Extensive experiments prove the high training efficiency, robustness, and state-of-the-art performance of our method on popular benchmarks. The code and the Inverse-Text test set are available at https://github.com/ymy-k/DPText-DETR.

The DETR-like segmentors have underpinned the most recent breakthroughs in semantic segmentation, which end-to-end train a set of queries representing the class prototypes or target segments. Recently, masked attention [8] is proposed to restrict each query to only attend to the foreground regions predicted by the preceding decoder block for easier optimization. Although promising, it relies on the learnable parameterized positional queries which tend to encode the dataset statistics, leading to inaccurate localization for distinct individual queries. In this paper, we propose a simple yet effective query design for semantic segmentation termed Dynamic Focus-aware Positional Queries (DFPQ), which dynamically generates positional queries conditioned on the cross-attention scores from the preceding decoder block and the positional encodings for the corresponding image features, simultaneously. Therefore, our DFPQ preserves rich localization information for the target segments and provides accurate and fine-grained positional priors. In addition, we propose to efficiently deal with high-resolution cross-attention by only aggregating the con-textual tokens based on the low-resolution cross-attention scores to perform local relation aggregation. Extensive experiments on ADE20K and Cityscapes show that with the two modifications on Mask2former, our framework achieves SOTA performance and outperforms Mask2former by clear margins of 1.1%, 1.9%, and 1.1% single-scale mIoU with ResNet-50, Swin-T, and Swin-B backbones on the ADE20K validation set, respectively. Source code is available at https://github.com/ziplab/FASeg.

Different from conventional image matting, which either requires user-defined scribbles/trimap to extract a specific foreground object or directly extracts all the foreground objects in the image indiscriminately, we introduce a new task named Referring Image Matting (RIM) in this paper, which aims to extract the meticulous alpha matte of the specific object that best matches the given natural language description, thus enabling a more natural and simpler instruction for image matting. First, we establish a large-scale challenging dataset Ref Matte by designing a comprehensive image composition and expression generation engine to automatically produce high-quality images along with diverse text attributes based on public datasets. RefMatte consists of 230 object categories, 47,500 images, 118,749 expression-region entities, and 474,996 expressions. Additionally, we construct a real-world test set with 100 high-resolution natural images and manually annotate complex phrases to evaluate the out-of-domain generalization abilities of RIM methods. Furthermore, we present a novel baseline method CLIPMat for RIM, including a context-embedded prompt, a text-driven semantic pop-up, and a multi-level details extractor. Extensive experiments on RefMatte in both keyword and expression settings validate the superiority of CLIPMat over representative methods. We hope this work could provide novel insights into image matting and encourage more follow-up studies. The dataset, code and models are available at https://github.com/JizhiziLi/RIM.

Perceiving potential “action possibilities” (i.e., affordance) regions of images and learning interactive functionalities of objects from human demonstration is a challenging task due to the diversity of human-object interactions. Prevailing affordance learning algorithms often adopt the label assignment paradigm and presume that there is a unique relationship between functional region and affordance label, yielding poor performance when adapting to unseen environments with large appearance variations. In this paper, we propose to leverage interactive affinity for affordance learning, i.e. extracting interactive affinity from human-object interaction and transferring it to non-interactive objects. Interactive affinity, which represents the contacts between different parts of the human body and local regions of the target object, can provide inherent cues of interconnectivity between humans and objects, thereby reducing the ambiguity of the perceived action possibilities. Specifically, we propose a pose-aided interactive affinity learning framework that exploits human pose to guide the network to learn the interactive affinity from human-object interactions. Particularly, a keypoint heuristic perception (KHP) scheme is devised to exploit the keypoint association of human pose to alleviate the uncertainties due to interaction diversities and contact occlusions. Besides, a contact-driven affordance learning (CAL) dataset is constructed by collecting and labeling over 5, 000 images. Experimental results demonstrate that our method outperforms the representative models regarding objective metrics and visual quality. Code and dataset: github.com/lhc1224/PIAL-Net.

End-to-end text spotting aims to integrate scene text detection and recognition into a unified framework. Dealing with the relationship between the two sub-tasks plays a pivotal role in designing effective spotters. Although Transformer-based methods eliminate the heuristic postprocessing, they still suffer from the synergy issue between the sub-tasks and low training efficiency. In this paper, we present DeepSolo, a simple DETR-like baseline that lets a single Decoder with Explicit Points Solo for text detection and recognition simultaneously. Technically, for each text instance, we represent the character sequence as ordered points and model them with learnable explicit point queries. After passing a single decoder, the point queries have encoded requisite text semantics and locations, thus can be further decoded to the center line, boundary, script, and confidence of text via very simple prediction heads in parallel. Besides, we also introduce a text-matching criterion to deliver more accurate supervisory signals, thus enabling more efficient training. Quantitative experiments on public benchmarks demonstrate that DeepSolo outperforms previous state-of-the-art methods and achieves better training efficiency. In addition, DeepSolo is also compatible with line annotations, which require much less annotation cost than polygons. The code is available at https://github.com/ViTAE-Transformer/DeepSolo.

Animal pose estimation is challenging for existing image-based methods because of limited training data and large intra- and inter-species variances. Motivated by the progress of visual-language research, we propose that pre-trained language models (e.g., CLIP) can facilitate animal pose estimation by providing rich prior knowledge for describing animal keypoints in text. However, we found that building effective connections between pre-trained language models and visual animal keypoints is non-trivial since the gap between text-based descriptions and keypoint-based visual features about animal pose can be significant. To address this issue, we introduce a novel prompt-based Contrastive learning scheme for connecting Language and AniMal Pose (CLAMP) effectively. The CLAMP attempts to bridge the gap by adapting the text prompts to the animal keypoints during network training. The adaptation is decomposed into spatialaware and feature-aware processes, and two novel contrastive losses are devised correspondingly. In practice, the CLAMP enables the first cross-modal animal pose estimation paradigm. Experimental results show that our method achieves state-of-the-art performance under the supervised, few-shot, and zero-shot settings, outperforming image-based methods by a large margin. The code is available at https://github.com/xuzhang1199/CLAMP.

This paper explores Deployment Authorization (DPA) as a means of restricting the generalization capabilities of vision models on certain domains to protect intellectual property. Nevertheless, the current advancements in DPA are predominantly confined to fully supervised settings. Such settings require the accessibility of annotated images from any unauthorized domain, rendering the DPA approaches impractical for real-world applications due to its exorbitant costs.To address this issue, we propose Source-Only Deployment Authorization (SDPA), which assumes that only authorized domains are accessible during training phases, and the model’s performance on unauthorized domains must be suppressed in inference stages. Drawing inspiration from distributional robust statistics, we present a lightweight method called Domain-Specified Optimization (DSO) for SDPA that degrades the model’s generalization over a divergence ball. DSO comes with theoretical guarantees on the convergence property and its authorization performance. As a complementary of SDPA, we also propose Target-Combined Deployment Authorization (TPDA), where unauthorized domains are partially accessible, and simplify the DSO method to a perturbation operation on the pseudo predictions, referred to as Target-Dependent Domain-Specified Optimization (TDSO). We demonstrate the effectiveness of our proposed DSO and TDSO methods through extensive experiments on six image benchmarks, achieving dominant performance on both SDPA and TDPA settings.

Visual Parameter-Efficient Fine-Tuning (PEFT) has become a powerful alternative for full fine-tuning so as to adapt pre-trained vision models to downstream tasks, which only tunes a small number of parameters while freezing the vast majority ones to ease storage burden and optimization difficulty. However, existing PEFT methods introduce trainable parameters to the same positions across different tasks depending solely on human heuristics and neglect the domain gaps. To this end, we study where to introduce and how to allocate trainable parameters by proposing a novel Sensitivity-aware visual Parameter-efficient fine-Tuning (SPT) scheme, which adaptively allocates trainable parameters to task-specific important positions given a desired tunable parameter budget. Specifically, our SPT first quickly identifies the sensitive parameters that require tuning for a given task in a data-dependent way. Next, our SPT further boosts the representational capability for the weight matrices whose number of sensitive parameters exceeds a pre-defined threshold by utilizing existing structured tuning methods, e.g., LoRA [21] or Adapter [20], to replace directly tuning the selected sensitive parameters (unstructured tuning) under the budget. Extensive experiments on a wide range of downstream recognition tasks show that our SPT is complementary to the existing PEFT methods and largely boosts their performance, e.g., SPT improves Adapter with supervised pre-trained ViT-B/16 backbone by 4.2% and 1.4% mean Top-1 accuracy, reaching SOTA performance on FGVC and VTAB-1k benchmarks, respectively. Source code is at https://github.com/ziplab/SPT.

Single hyperspectral image super-resolution (single-HSI-SR) aims to restore a high-resolution hyperspectral image from a low-resolution observation. However, the prevailing CNN-based approaches have shown limitations in building long-range dependencies and capturing interaction information between spectral features. This results in inadequate utilization of spectral information and artifacts after upsampling. To address this issue, we propose ES-SAformer, an ESSA attention-embedded Transformer network for single-HSI-SR with an iterative refining structure. Specifically, we first introduce a robust and spectral-friendly similarity metric, i.e., the spectral correlation coefficient of the spectrum (SCC), to replace the original attention matrix and incorporates inductive biases into the model to facilitate training. Built upon it, we further utilize the kernelizable attention technique with theoretical support to form a novel efficient SCC-kernel-based self-attention (ESSA) and reduce attention computation to linear complexity. ESSA enlarges the receptive field for features after upsampling without bringing much computation and allows the model to effectively utilize spatial-spectral information from different scales, resulting in the generation of more natural high-resolution images. Without the need for pretraining on large-scale datasets, our experiments demonstrate ESSA’s effectiveness in both visual quality and quantitative results. The code will be released at ESSAformer.

Two-stage point-to-box network acts as a critical role in the recent popular 3D Siamese tracking paradigm, which first generates proposals and then predicts corresponding proposal-wise scores. However, such a network suffers from tedious hyper-parameter tuning and task misalignment, limiting the tracking performance. Towards these concerns, we propose a simple yet effective one-stage point-to-box network for point cloud-based 3D single object tracking. It synchronizes 3D proposal generation and center-ness score prediction by a parallel predictor without tedious hyper-parameters. To guide a task-aligned score ranking of proposals, a center-aware focal loss is proposed to supervise the training of the center-ness branch, which enhances the network's discriminative ability to distinguish proposals of different quality. Besides, we design a binary target classifier to identify target-relevant points. By integrating the derived classification scores with the center-ness scores, the resulting network can effectively suppress interference proposals and further mitigate task misalignment. Finally, we present a novel one-stage Siamese tracker OSP2B equipped with the designed network. Extensive experiments on challenging benchmarks including KITTI and Waymo SOT Dataset show that our OSP2B achieves leading performance with a considerable real-time speed.

Although human reconstruction typically results in human-specific avatars, recent 3D scene reconstruction techniques utilizing pixel-aligned features show promise in generalizing to new scenes. Applying these techniques to human avatar reconstruction can result in a volumetric avatar with generalizability but limited animatability due to rendering only being possible for static representations. In this paper, we propose AniPixel, a novel animatable and generalizable human avatar reconstruction method that leverages pixel-aligned features for body geometry prediction and RGB color blending. Technically, to align the canonical space with the target space and the observation space, we propose a bidirectional neural skinning field based on skeleton-driven deformation to establish the target-to-canonical and canonical-to-observation correspondences. Then, we disentangle the canonical body geometry into a normalized neutral-sized body and a subject-specific residual for better generalizability. As the geometry and appearance are closely related, we introduce pixel-aligned features to facilitate the body geometry prediction and detailed surface normals to reinforce the RGB color blending. We also devise a pose-dependent and view direction-related shading module to represent the local illumination variance. Experiments show that AniPixel renders comparable novel views while delivering better novel pose animation results than state-of-the-art methods. Code will be released at AniPixel. https://github.com/loong8888/AniPixel.

Demystifying complex human-ground interactions is essential for accurate and realistic 3D human motion reconstruction from RGB videos, as it ensures consistency between the humans and the ground plane. Prior methods have modeled human-ground interactions either implicitly or in a sparse manner, often resulting in unrealistic and incorrect motions when faced with noise and uncertainty. In contrast, our approach explicitly represents these interactions in a dense and continuous manner. To this end, we propose a novel Ground-aware Motion Model for 3D Human Motion Reconstruction, named GraMMaR, which jointly learns the distribution of transitions in both pose and interaction between every joint and ground plane at each time step of a motion sequence. It is trained to explicitly promote consistency between the motion and distance change towards the ground. After training, we establish a joint optimization strategy that utilizes GraMMaR as a dual-prior, regularizing the optimization towards the space of plausible ground-aware motions. This leads to realistic and coherent motion reconstruction, irrespective of the assumed or learned ground plane. Through extensive evaluation on the AMASS and AIST++ datasets, our model demonstrates good generalization and discriminating abilities in challenging cases including complex and ambiguous human-ground interactions. The code will be available at https://github.com/xymsh/GraMMaR.

The success of the Segment Anything Model (SAM) demonstrates the significance of data-centric machine learning. However, due to the difficulties and high costs associated with annotating Remote Sensing (RS) images, a large amount of valuable RS data remains unlabeled, particularly at the pixel level. In this study, we leverage SAM and existing RS object detection datasets to develop an efficient pipeline for generating a large-scale RS segmentation dataset, dubbed SAMRS. SAMRS totally possesses 105,090 images and 1,668,241 instances, surpassing existing high-resolution RS segmentation datasets in size by several orders of magnitude. It provides object category, location, and instance information that can be used for semantic segmentation, instance segmentation, and object detection, either individually or in combination. We also provide a comprehensive analysis of SAMRS from various aspects. Moreover, preliminary experiments highlight the importance of conducting segmentation pre-training with SAMRS to address task discrepancies and alleviate the limitations posed by limited training data during fine-tuning. The code and dataset will be available at https://github.com/ViTAE-Transformer/SAMRS

Abstract Current object detectors typically have a feature pyramid (FP) module for multi-level feature fusion (MFF) which aims to mitigate the gap between features from different levels and form a comprehensive object representation to achieve better detection performance. However, they usually require heavy cross-level connections or iterative refinement to obtain better MFF result, making them complicated in structure and inefficient in computation. To address these issues, we propose a novel and efficient context modeling mechanism that can help existing FPs deliver better MFF results while reducing the computational costs effectively. In particular, we introduce a novel insight that comprehensive contexts can be decomposed and condensed into two types of representations for higher efficiency. The two representations include a locally concentrated representation and a globally summarized representation, where the former focuses on extracting context cues from nearby areas while the latter extracts general contextual representations of the whole image scene as global context cues. By collecting the condensed contexts, we employ a Transformer decoder to investigate the relations between them and each local feature from the FP and then refine the MFF results accordingly. As a result, we obtain a simple and light-weight Transformer-based Context Condensation (TCC) module, which can boost various FPs and lower their computational costs simultaneously. Extensive experimental results on the challenging MS COCO dataset show that TCC is compatible to four representative FPs and consistently improves their detection accuracy by up to 7.8% in terms of average precision and reduce their complexities by up to around 20% in terms of GFLOPs, helping them achieve state-of-the-art performance more efficiently. Code will be released at https://github.com/zhechen/TCC .

Recently, there has been an increasing concern about the privacy issue raised by identifiable information in machine learning. However, previous portrait matting methods were all based on identifiable images. To fill the gap, we present P3M-10k, which is the first large-scale anonymized benchmark for Privacy-Preserving Portrait Matting (P3M). P3M-10k consists of 10,421 high resolution face-blurred portrait images along with high-quality alpha mattes, which enables us to systematically evaluate both trimap-free and trimap-based matting methods and obtain some useful findings about model generalization ability under the privacy preserving training (PPT) setting. We also present a unified matting model dubbed P3M-Net that is compatible with both CNN and transformer backbones. To further mitigate the cross-domain performance gap issue under the PPT setting, we devise a simple yet effective Copy and Paste strategy (P3M-CP), which borrows facial information from public celebrity images and directs the network to reacquire the face context at both data and feature level. Extensive experiments on P3M-10k and public benchmarks demonstrate the superiority of P3M-Net over state-of-the-art methods and the effectiveness of P3M-CP in improving the cross-domain generalization ability, implying a great significance of P3M for future research and real-world applications. The dataset, code and models are available here (https://github.com/ViTAE-Transformer/P3M-Net).

Abstract Recent studies have shown promising results on joint learning of local feature detectors and descriptors. To address the lack of ground-truth keypoint supervision, previous methods mainly inject appropriate knowledge about keypoint attributes into the network to facilitate model learning. In this paper, inspired by traditional corner detectors, we develop an end-to-end deep network, named Deep Corner, which adds a local similarity-based keypoint measure into a plain convolutional network. Deep Corner enables finding reliable keypoints and thus benefits the learning of the distinctive descriptors. Moreover, to improve keypoint localization, we first study previous multi-level keypoint detection strategies and then develop a multi-level U-Net architecture, where the similarity of features at multiple levels can be exploited effectively. Finally, to improve the invariance of descriptors, we propose a feature self-transformation operation, which transforms the learned features adaptively according to the specific local information. The experimental results on several tasks and comprehensive ablation studies demonstrate the effectiveness of our method and the involved components.

Image complexity (IC) is an essential visual perception for human beings to understand an image. However, explicitly evaluating the IC is challenging, and has long been overlooked since, on the one hand, the evaluation of IC is relatively subjective due to its dependence on human perception, and on the other hand, the IC is semantic-dependent while real-world images are diverse. To facilitate the research of IC assessment in this deep learning era, we built the first, to our best knowledge, large-scale IC dataset with 9,600 well-annotated images. The images are of diverse areas such as abstract, paintings and real-world scenes, each of which is elaborately annotated by 17 human contributors. Powered by this high-quality dataset, we further provide a base model to predict the IC scores and estimate the complexity density maps in a weakly supervised way. The model is verified to be effective, and correlates well with human perception (with the Pearson correlation coefficient being 0.949). Last but not the least, we have empirically validated that the exploration of IC can provide auxiliary information and boost the performance of a wide range of computer vision tasks. The dataset and source code can be found at https://github.com/tinglyfeng/IC9600.

Previous human parsing methods are limited to parsing humans into pre-defined classes, which is inflexible for practical fashion applications that often have new fashion item classes. In this paper, we define a novel one-shot human parsing (OSHP) task that requires parsing humans into an open set of classes defined by any test example. During training, only base classes are exposed, which only overlap with part of the test-time classes. To address three main challenges in OSHP, i.e., small sizes, testing bias, and similar parts, we devise an End-to-end One-shot human Parsing Network (EOP-Net). Firstly, an end-to-end human parsing framework is proposed to parse the query image into both coarse-grained and fine-grained human classes, which embeds rich semantic information that is shared across different granularities to identify the small-sized human classes. Then, we gradually smooth the training-time static prototypes to get robust class representations. Moreover, we employ a dynamic objective to encourage the network enhancing features' representational capability in the early training phase while improving features' transferability in the late training phase. Therefore, our method can quickly adapt to the novel classes and mitigate the testing bias issue. In addition, we add a contrastive loss at the prototype level to enforce inter-class distances, thereby discriminating the similar parts. For comprehensive evaluations on the new task, we tailor three existing popular human parsing benchmarks to the OSHP task. Experiments demonstrate that EOP-Net outperforms representative one-shot segmentation models by large margins and serves as a strong baseline for further research.

We present a unified formulation and model for three motion and 3D perception tasks: optical flow, rectified stereo matching and unrectified stereo depth estimation from posed images. Unlike previous specialized architectures for each specific task, we formulate all three tasks as a unified dense correspondence matching problem, which can be solved with a single model by directly comparing feature similarities. Such a formulation calls for discriminative feature representations, which we achieve using a Transformer, in particular the cross-attention mechanism. We demonstrate that cross-attention enables integration of knowledge from another image via cross-view interactions, which greatly improves the quality of the extracted features. Our unified model naturally enables cross-task transfer since the model architecture and parameters are shared across tasks. We outperform RAFT with our unified model on the challenging Sintel dataset, and our final model that uses a few additional task-specific refinement steps outperforms or compares favorably to recent state-of-the-art methods on 10 popular flow, stereo and depth datasets, while being simpler and more efficient in terms of model design and inference speed.

Most existing convolutional neural-network-based super-resolution (SR) methods focus on designing effective neural blocks but rarely describe the image SR mechanism from the perspective of image evolution in the SR process. In this study, we explore a new research routine by abstracting the movement of pixels in the reconstruction process as the flow of fluid in the field of fluid dynamics (FD), where explicit motion laws of particles have been discovered. Specifically, a novel fluid micelle network is devised for image SR based on the theory of FD that follows the residual learning scheme but learns the residual structure by solving the finite difference equation in FD. The pixel motion equation in the SR process is derived from the Navier-Stokes (N-S) FD equation, establishing a guided branch that is aware of edge information. Thus, the second-order residual drives the network for feature extraction, and the guided branch corrects the direction of the pixel stream to supplement the details. Experiments on popular benchmarks and a real-world microscope chip image dataset demonstrate that the proposed method outperforms other modern methods in terms of both objective metrics and visual quality. The proposed method can also reconstruct clear geometric structures, offering the potential for real-world applications.

Deep learning has largely reshaped remote sensing (RS) research for aerial image understanding and made a great success. Nevertheless, most of the existing deep models are initialized with the ImageNet pretrained weights since natural images inevitably present a large domain gap relative to aerial images, probably limiting the fine-tuning performance on downstream aerial scene tasks. This issue motivates us to conduct an empirical study of RS pretraining (RSP) on aerial images. To this end, we train different networks from scratch with the help of the largest RS scene recognition dataset up to now—MillionAID—to obtain a series of RS pretrained backbones, including both convolutional neural networks (CNNs) and vision transformers, such as Swin and ViTAE, which have shown promising performance on computer vision tasks. Then, we investigate the impact of RSP on representative downstream tasks, including scene recognition, semantic segmentation, object detection, and change detection using these CNN and vision transformer backbones. Empirical study shows that RSP can help deliver distinctive performances in scene recognition tasks and in perceiving RS-related semantics, such as “Bridge” and “Airplane.” We also find that, although RSP mitigates the data discrepancies of traditional ImageNet pretraining on RS images, it may still suffer from task discrepancies, where downstream tasks require different representations from scene recognition tasks. These findings call for further research efforts on both large-scale pretraining datasets and effective pretraining methods. The codes and pretrained models will be released athttps://github.com/ViTAE-Transformer/ViTAE-Transformer-Remote-Sensing.

Large-scale vision foundation models have made significant progress in visual tasks on natural images, with vision transformers (ViTs) being the primary choice due to their good scalability and representation ability. However, large-scale models in remote sensing (RS) have not yet been sufficiently explored. In this article, we resort to plain ViTs with about 100 million parameters and make the first attempt to propose large vision models tailored to RS tasks and investigate how such large models perform. To handle the large sizes and objects of arbitrary orientations in RS images, we propose a new rotated varied-size window attention to replace the original full attention in transformers, which can significantly reduce the computational cost and memory footprint while learning better object representation by extracting rich context from the generated diverse windows. Experiments on detection tasks show the superiority of our model over all state-of-the-art models, achieving 81.24% mean average precision (mAP) on the DOTA-V1.0 dataset. The results of our models on downstream classification and segmentation tasks also show competitive performance compared to existing advanced methods. Further experiments show the advantages of our models in terms of computational complexity and data efficiency in transferring. The code and models will be released athttps://github.com/ViTAE-Transformer/Remote-Sensing-RVSA.

Infrared small target detection (IRSTD) is important for many practical applications such as hazardous aircraft warning, especially when the target is not visible in visible light image due to atmospheric conditions such as fog and cloud. However, IRSTD is challenging due to noises, small and dim targets. To address this challenge, we propose a novel Dim2Clear Network (Dim2Clear) for IRSTD in this paper. Specifically, the Dim2Clear consists of a U-Net backbone encoder, a context mixer decoder (CMD) based on spatial and frequency attention (SFA), and an eyeball-shaped enhancement module (EEM). The CMD is composed of cascaded regular residual blocks where two SFA modules are inserted. Each SFA module receives features from different residual blocks and generates spatial attention map from them to modulate the low-level features, which are then decomposed into low and high frequencies using the discrete cosine transformation. Accordingly, features are further modulated according to the generated frequency attention maps. In this way, SFA can extract both spatial context and frequency context to improve the feature representation capacity. In addition, we design an EEM to suppress the noise and enhance the signal-to-noise ratio in the segmentation results from the perspective of image super-resolution. Experiments on the SIRST dataset and our newly constructed IRSTD-1k dataset show that the proposed Dim2Clear outperforms state-of-the-art methods.

Deep learning has demonstrated its power in image rectification by leveraging the representation capacity of deep neural networks via supervised training based on a large-scale synthetic dataset. However, the model may overfit the synthetic images and generalize not well on real-world fisheye images due to the limited universality of a specific distortion model and the lack of explicitly modeling the distortion and rectification process. In this paper, we propose a novel self-supervised image rectification (SIR) method based on an important insight that the rectified results of distorted images of a same scene from different lenses should be the same. Specifically, we devise a new network architecture with a shared encoder and several prediction heads, each of which predicts the distortion parameter of a specific distortion model. We further leverage a differentiable warping module to generate the rectified images and re-distorted images from the distortion parameters and exploit the intra- and inter-model consistency between them during training, thereby leading to a self-supervised learning scheme without the need for ground-truth distortion parameters or normal images. Experiments on synthetic dataset and real-world fisheye images demonstrate that our method achieves comparable or even better performance than the supervised baseline method and representative state-of-the-art (SOTA) methods. The proposed self-supervised method also provides a possible way to improve the universality of distortion models while keeping their self-consistency. Code and datasets will be available at https://github.com/loong8888/SIR.

Unsupervised person re-identification is a challenging and promising task in computer vision. Nowadays unsupervised person re-identification methods have achieved great progress by training with pseudo labels. However, how to purify feature and label noise is less explicitly studied in the unsupervised manner. To purify the feature, we take into account two types of additional features from different local views to enrich the feature representation. The proposed multi-view features are carefully integrated into our cluster contrast learning to leverage more discriminative cues that the global feature easily ignored and biased. To purify the label noise, we propose to take advantage of the knowledge of teacher model in an offline scheme. Specifically, we first train a teacher model from noisy pseudo labels, and then use the teacher model to guide the learning of our student model. In our setting, the student model could converge fast with the supervision of the teacher model thus reduce the interference of noisy labels as the teacher model greatly suffered. After carefully handling the noise and bias in the feature learning, our purification modules are proven to be very effective for unsupervised person re-identification. Extensive experiments on two popular person re-identification datasets demonstrate the superiority of our method. Especially, our approach achieves a state-of-the-art accuracy 85.8% @mAP and 94.5% @Rank-1 on the challenging Market-1501 benchmark with ResNet-50 under the fully unsupervised setting. Code has been available at: https://github.com/tengxiao14/Purification_ReID.

Hyperspectral image (HSI) classification is challenging due to spatial variability caused by complex imaging conditions. Prior methods suffer from limited representation ability, as they train specially designed networks from scratch on limited annotated data. We propose a tri-spectral image generation pipeline that transforms HSI into high-quality tri-spectral images, enabling the use of off-the-shelf ImageNet pretrained backbone networks for feature extraction. Motivated by the observation that there are many homogeneous areas with distinguished semantic and geometric properties in HSIs, which can be used to extract useful contexts, we propose an end-to-end segmentation network named DCN-T. It adopts transformers to effectively encode regional adaptation and global aggregation spatial contexts within and between the homogeneous areas discovered by similarity-based clustering. To fully exploit the rich spectrums of the HSI, we adopt an ensemble approach where all segmentation results of the tri-spectral images are integrated into the final prediction through a voting scheme. Extensive experiments on three public benchmarks show that our proposed method outperforms state-of-the-art methods for HSI classification. The code will be released at https://github.com/DotWang/DCN-T.

Detecting hardware Trojan (HT) from a microscope chip image (MCI) is crucial for many applications, such as financial infrastructure and transport security. It takes an inordinate cost in scanning high-resolution (HR) microscope images for HT detection. It is useful when the chip image is in low-resolution (LR), which can be acquired faster and at a lower cost than its HR counterpart. However, the lost details and noises due to the electric charge effect in LR MCIs will affect the detection performance, making the problem more challenging. In this article, we address this issue by first discussing why recovering curvature information matters for HT detection and then proposing a novel MCI super-resolution (SR) method via a curvature consistent network (CCN). It consists of a homogeneous workflow and a heterogeneous workflow, where the former learns a mapping between homogeneous images, i.e., LR and HR MCIs, and the latter learns a mapping between heterogeneous images, i.e., MCIs and curvature images. Besides, a collaborative fusion strategy is used to leverage features learned from both workflows level-by-level by recovering the HR image eventually. To mitigate the issue of lacking an MCI dataset, we construct a new benchmark consisting of realistic MCIs at different resolutions, called MCI. Experiments on MCI demonstrate that the proposed CCN outperforms representative SR methods by recovering more delicate circuit lines and yields higher HT detection performance. The dataset is available at github.com/RuiZhang97/CCN.

Although point-based networks are demonstrated to be accurate for 3D point cloud modeling, they are still falling behind their voxel-based competitors in 3D detection. We observe that the prevailing set abstraction design for down-sampling points may maintain too much unimportant background information that can affect feature learning for detecting objects. To tackle this issue, we propose a novel set abstraction method named Semantics-Augmented Set Abstraction (SASA). Technically, we first add a binary segmentation module as the side output to help identify foreground points. Based on the estimated point-wise foreground scores, we then propose a semantics-guided point sampling algorithm to help retain more important foreground points during down-sampling. In practice, SASA shows to be effective in identifying valuable points related to foreground objects and improving feature learning for point-based 3D detection. Additionally, it is an easy-to-plug-in module and able to boost various point-based detectors, including single-stage and two-stage ones. Extensive experiments on the popular KITTI and nuScenes datasets validate the superiority of SASA, lifting point-based detection models to reach comparable performance to state-of-the-art voxel-based methods. Code is available at https://github.com/blakechen97/SASA.

Existing Scene Text Recognition (STR) methods typically use a language model to optimize the joint probability of the 1D character sequence predicted by a visual recognition (VR) model, which ignore the 2D spatial context of visual semantics within and between character instances, making them not generalize well to arbitrary shape scene text. To address this issue, we make the first attempt to perform textual reasoning based on visual semantics in this paper. Technically, given the character segmentation maps predicted by a VR model, we construct a subgraph for each instance, where nodes represent the pixels in it and edges are added between nodes based on their spatial similarity. Then, these subgraphs are sequentially connected by their root nodes and merged into a complete graph. Based on this graph, we devise a graph convolutional network for textual reasoning (GTR) by supervising it with a cross-entropy loss. GTR can be easily plugged in representative STR models to improve their performance owing to better textual reasoning. Specifically, we construct our model, namely S-GTR, by paralleling GTR to the language model in a segmentation-based STR baseline, which can effectively exploit the visual-linguistic complementarity via mutual learning. S-GTR sets new state-of-the-art on six challenging STR benchmarks and generalizes well to multi-linguistic datasets. Code is available at https://github.com/adeline-cs/GTR.

Semi-supervised video object segmentation (VOS) refers to segmenting the target object in remaining frames given its annotation in the first frame, which has been actively studied in recent years. The key challenge lies in finding effective ways to exploit the spatio-temporal context of past frames to help learn discriminative target representation of current frame. In this paper, we propose a novel Siamese network with a specifically designed interactive transformer, called SITVOS, to enable effective context propagation from historical to current frames. Technically, we use the transformer encoder and decoder to handle the past frames and current frame separately, i.e., the encoder encodes robust spatio-temporal context of target object from the past frames, while the decoder takes the feature embedding of current frame as the query to retrieve the target from the encoder output. To further enhance the target representation, a feature interaction module (FIM) is devised to promote the information flow between the encoder and decoder. Moreover, we employ the Siamese architecture to extract backbone features of both past and current frames, which enables feature reuse and is more efficient than existing methods. Experimental results on three challenging benchmarks validate the superiority of SITVOS over state-of-the-art methods. Code is available at https://github.com/LANMNG/SITVOS.

Estimating Click-Through Rate (CTR) is a vital yet challenging task in personalized product search. However, existing CTR methods still struggle in the product search settings due to the following three challenges including how to more effectively extract users' short-term interest with respect to multiple aspects, how to extract and fuse users' long-term interest with short-term interest, how to address the entangling characteristic of long and short-term interests. To resolve these challenges, in this paper, we propose a new approach named Hierarchical Interests Fusing Network (HIFN), which consists of four basic modules namely Short-term Interest Extractor (SIE), Long-term Interest Extractor (LIE), Interest Fusion Module (IFM) and Interest Disentanglement Module (IDM). Specifically, SIE is proposed to extract user's short-term interest by integrating three fundamental interest encoders within it namely query-dependent, target-dependent and causal-dependent interest encoder, respectively, followed by delivering the resultant representation to the module LIE, where it can effectively capture user long-term interest by devising an attention mechanism with respect to the short-term interest from SIE module. In IFM, the achieved long and short-term interests are further fused in an adaptive manner, followed by concatenating it with original raw context features for the final prediction result. Last but not least, considering the entangling characteristic of long and short-term interests, IDM further devises a self-supervised framework to disentangle long- and short-term interests. Extensive offline and online evaluations on a real-world e-commerce platform demonstrate the superiority of HIFN over state-of-the-art methods.

Transformers are successfully applied to computer vision due to their powerful modeling capacity with self-attention. However, the excellent performance of transformers heavily depends on enormous training images. Thus, a data-efficient transformer solution is urgently needed. In this work, we propose an early knowledge distillation framework, which is termed as DearKD, to improve the data efficiency required by transformers. Our DearKD is a two-stage framework that first distills the inductive biases from the early intermediate layers of a CNN and then gives the transformer full play by training without distillation. Further, our DearKD can be readily applied to the extreme data-free case where no real images are available. In this case, we propose a boundary-preserving intra-divergence loss based on DeepInversion to further close the performance gap against the full-data counterpart. Extensive experiments on ImageNet, partial ImageNet, data-free setting and other downstream tasks prove the superiority of DearKD over its baselines and state-of-the-art methods.

Although detection with Transformer (DETR) is increasingly popular, its global attention modeling requires an extremely long training period to optimize and achieve promising detection performance. Alternative to existing studies that mainly develop advanced feature or embedding designs to tackle the training issue, we point out that the Region-of-Interest (RoI) based detection refinement can easily help mitigate the difficulty of training for DETR methods. Based on this, we introduce a novel REcurrent Glimpse-based decOder (REGO) in this paper. In particular, the REGO employs a multi-stage recurrent processing structure to help the attention of DETR gradually focus on foreground objects more accurately. In each processing stage, visual features are extracted as glimpse features from RoIs with enlarged bounding box areas of detection results from the previous stage. Then, a glimpse-based decoder is introduced to provide refined detection results based on both the glimpse features and the attention modeling outputs of the previous stage. In practice, REGO can be easily embedded in representative DETR variants while maintaining their fully end-to-end training and inference pipelines. In particular, REGO helps Deformable DETR achieve 44.8 AP on the MSCOCO dataset with only 36 training epochs, compared with the first DETR and the Deformable DETR that require 500 and 50 epochs to achieve comparable performance, respectively. Experiments also show that REGO consistently boosts the performance of different DETR detectors by up to 7% relative gain at the same setting of 50 training epochs. Code is available via https://github.com/zhechen/Deformable-DETR-REGO.

In recent years, the security of AI systems has drawn increasing research attention, especially in the medical imaging realm. To develop a secure medical image analysis (MIA) system, it is a must to study possible backdoor attacks (BAs), which can embed hidden malicious behaviors into the system. However, designing a unified BA method that can be applied to various MIA systems is challenging due to the diversity of imaging modalities (e.g., X-Ray, CT, and MRI) and analysis tasks (e.g., classification, detection, and segmentation). Most existing BA methods are designed to attack natural image classification models, which apply spatial triggers to training images and inevitably corrupt the semantics of poisoned pixels, leading to the failures of attacking dense prediction models. To address this issue, we propose a novel Frequency-Injection based Backdoor Attack method (FIBA) that is capable of delivering attacks in various MIA tasks. Specifically, FIBA leverages a trigger function in the frequency domain that can inject the low-frequency information of a trigger image into the poisoned image by linearly combining the spectral amplitude of both images. Since it preserves the semantics of the poisoned image pixels, FIBA can perform attacks on both classification and dense prediction models. Experiments on three benchmarks in MIA (i.e., ISIC-2019 [4] for skin lesion classification, KiTS-19 [17] for kidney tumor segmentation, and EAD-2019 [1] for endoscopic artifact detection), validate the effectiveness of FIBA and its superiority over stateof-the-art methods in attacking MIA models and bypassing backdoor defense. Source code will be available at code.

Scene graph generation (SGG) aims to detect objects and predict the relationships between each pair of objects. Existing SGG methods usually suffer from several issues, including 1) ambiguous object representations, as graph neural network-based message passing (GMP) modules are typically sensitive to spurious inter-node correlations, and 2) low diversity in relationship predictions due to severe class imbalance and a large number of missing annotations. To address both problems, in this paper, we propose a regu-larized unrolling network (RU-Net). We first study the relation between GMP and graph Laplacian denoising (GLD) from the perspective of the unrolling technique, determining that GMP can be formulated as a solver for GLD. Based on this observation, we propose an unrolled message passing module and introduce an fp-based graph regularization to suppress spurious connections between nodes. Second, we propose a group diversity enhancement module that pro-motes the prediction diversity of relationships via rank max-imization. Systematic experiments demonstrate that RU-Net is effective under a variety of settings and metrics. Fur-thermore, RU-Net achieves new state-of-the-arts on three popular databases: VG, VRD, and OI. Code is available at https://github.com/siml3/RU-Net.

Affordance grounding, a task to ground (i.e., localize) action possibility region in objects, which faces the challenge of establishing an explicit link with object parts due to the diversity of interactive affordance. Human has the ability that transform the various exocentric interactions to invariant egocentric affordance so as to counter the impact of interactive diversity. To empower an agent with such ability, this paper proposes a task of affordance grounding from exocentric view, i.e., given exocentric human-object interaction and egocentric object images, learning the affordance knowledge of the object and transferring it to the egocentric image using only the affordance label as supervision. To this end, we devise a cross-view knowledge transfer framework that extracts affordance-specific features from exocentric interactions and enhances the perception of affordance regions by preserving affordance correlation. Specifically, an Affordance Invariance Mining module is devised to extract specific clues by minimizing the intra-class differences originated from interaction habits in exocentric images. Besides, an Affordance Co-relation Preserving strategy is presented to perceive and localize affordance by aligning the co-relation matrix of predicted results between the two views. Particularly, an affordance grounding dataset named AGD20K is constructed by collecting and labeling over 20K images from 36 affordance categories. Experimental results demonstrate that our method outperforms the representative models in terms of objective metrics and visual quality. Code: github.com/lhc1224/Cross-View-AG.

Learning-based optical flow estimation has been dominated with the pipeline of cost volume with convolutions for flow regression, which is inherently limited to local correlations and thus is hard to address the long-standing challenge of large displacements. To alleviate this, the state-of-the-art framework RAFT gradually improves its prediction quality by using a large number of iterative refinements, achieving remarkable performance but introducing linearly increasing inference time. To enable both high accuracy and efficiency, we completely revamp the dominant flow regression pipeline by reformulating optical flow as a global matching problem, which identifies the correspondences by directly comparing feature similarities. Specifically, we propose a GMFlow framework, which consists of three main components: a customized Transformer for feature enhancement, a correlation and softmax layer for global feature matching, and a self-attention layer for flow propagation. We further introduce a refinement step that reuses GMFlow at higher feature resolution for residual flow prediction. Our new framework outperforms 31-refinements RAFT on the challenging Sintel benchmark, while using only one refinement and running faster, suggesting a new paradigm for accurate and efficient optical flow estimation. Code is available at https://github.com/haofeixu/gmflow.

Infrared small target detection (IRSTD) refers to extracting small and dim targets from blurred backgrounds, which has a wide range of applications such as traffic management and marine rescue. Due to the low signal-to-noise ratio and low contrast, infrared targets are easily submerged in the background of heavy noise and clutter. How to detect the precise shape information of infrared targets remains challenging. In this paper, we propose a novel infrared shape network (ISNet), where Taylor finite difference (TFD) -inspired edge block and two-orientation attention aggregation (TOAA) block are devised to address this problem. Specifically, TFD-inspired edge block aggregates and enhances the comprehensive edge information from different levels, in order to improve the contrast between target and background and also lay a foundation for extracting shape information with mathematical interpretation. TOAA block calculates the lowlevel information with attention mechanism in both row and column directions and fuses it with the high-level information to capture the shape characteristic of targets and suppress noises. In addition, we construct a new benchmark consisting of 1, 000 realistic images in various target shapes, different target sizes, and rich clutter backgrounds with accurate pixel-level annotations, called IRSTD-1k. Experiments on public datasets and IRSTD-1 k demonstrate the superiority of our approach over representative state-of-the-art IRSTD methods. The dataset and code are available at github.com/RuiZhang97/ISNet.

Monocular visual odometry (VO) has attracted extensive research attention by providing real-time vehicle motion from cost-effective camera images. However, state-of-the-art optimization-based monocular VO methods suffer from the scale inconsistency problem for long-term predictions. Deep learning has recently been introduced to address this issue by leveraging stereo sequences or ground-truth motions in the training dataset. However, it comes at an additional cost for data collection, and such training data may not be available in all datasets. In this work, we propose VRVO, a novel framework for retrieving the absolute scale from virtual data that can be easily obtained from modern simulation environments, whereas in the real domain no stereo or ground-truth data are required in either the training or inference phases. Specifically, we first train a scale-aware disparity network using both monocular real images and stereo virtual data. The virtual-to-real domain gap is bridged by using an adversarial training strategy to map images from both domains into a shared feature space. The resulting scale-consistent disparities are then integrated with a direct VO system by constructing a virtual stereo objective that ensures the scale consistency over long trajectories. Additionally, to address the suboptimality issue caused by the separate optimization backend and the learning process, we further propose a mutual reinforcement pipeline that allows bidirectional information flow between learning and optimization, which boosts the robustness and accuracy of each other. We demonstrate the effectiveness of our framework on the KITTI and vKITTI2 datasets.

Remote sensing image spatiotemporal fusion (STF) algorism plays an important role by supplementing the lack of original high-resolution remote sensing satellite images in the study scenarios of dense time-series data. In recent years, the deep-learning-based STF algorithm has become a research hotspot with comparatively higher accuracy and robustness. However, due to the lack of sufficient high-quality images for training and the huge resolution gap between low-resolution images and high-resolution images, it is difficult to recover detailed information, especially for areas of land-cover change. In this paper, we propose a two-layers super-resolution based generation adversarial spatiotemporal fusion model(TLSRSTF) using smaller inputs to reduce pressure on data requirements and a mutual affine convolution to reduce model parameters. Specifically, we only use a pair of high-resolution and low-resolution images and a high-resolution image at any time. A spatial degradation consistency is constructed to adaptively determine the ratio of two layers of the super-resolution STF model. The quantitative and qualitative experimental results on public spatiotemporal fusion datasets demonstrate our superiority over the state-of-the-art methods.

Synthetic Aperture Radar (SAR) becomes prevailing in remote sensing while SAR images are challenging to interpret by human visual perception due to the active imaging mechanism and speckle noise. Recent researches on SAR-to-optical image translation provide a promising solution and have attracted increasing attentions, though still suffering from low optical image quality with geometric distortion due to the large domain gap. In this paper, we mitigate this issue from a novel perspective, i.e., neural partial differential equations (PDE). First, based on the efficient numerical scheme for solving PDE, i.e., Taylor Central Difference (TCD), we devise a basic TCD residual block to build the backbone network, which promotes the extraction of useful information in SAR images by aggregating and enhancing features from different levels. Furthermore, inspired by the Perona-Malik Diffusion (PMD), we devise a PMD neural module to implement feature diffusion through layers, aiming at removing the noises in smooth regions while preserving the geometric structures. Assembling them together, we propose a novel SAR-to-Optical image translation network named S2O-NPDE, which delivers optical images with finer structures and less noise while enjoying an explainability advantage from explicit mathematical derivation. Experiments on the popular SEN1-2 dataset show that our model outperforms state-of-the-art methods in terms of both objective metrics and visual quality.

Given the mannequin (i.e., reference person) and target garment, the virtual try-on (VTON) task aims at dressing the mannequin in the provided garment automatically, having attracted increasing attention in recent years. Previous works usually conduct the garment deformation under the guidance of ''shape''. However, ''shape-only transformation'' ignores the local structures and results in unnatural distortions. To address this issue, we propose a Gated Try-on method by learning the ManneqUin-Specific Transformation (GT-MUST). Technically, we implement GT-MUST as a three-stage deep neural model. First, GT-MUST learns the ''mannequin-specific transformation'' with a ''take-off'' mechanism, which recovers the warped clothes of the mannequin to its original in-shop state. Then, the learned ''mannequin-specific transformation'' is inverted and utilized to help generate the mannequin-specific warped state for a target garment. Finally, a special gate is employed to better combine the mannequin-specific warped garment with the mannequin. GT-MUST benefits from learning to solve a much easier ''take-off'' task to obtain the mannequin-specific information than the common ''try-on'' task, since flat in-shop garments usually have less variation in shape than those clothed on the body. Experiments on the fashion dataset demonstrate that GT-MUST outperforms the state-of-the-art virtual try-on methods. The code is available at https://github.com/wangning-001/GT-MUST.

Infrared small target detection (IRSTD) refers to segmenting the small targets from infrared images, which is of great significance in practical applications. However, due to the small scale of targets as well as noise and clutter in the background, current deep neural network-based methods struggle in extracting features with discriminative semantics while preserving fine details. In this paper, we address this problem by proposing a novel RKformer model with an encoder-decoder structure, where four specifically designed Runge-Kutta transformer (RKT) blocks are stacked sequentially in the encoder. Technically, it has three key designs. First, we adopt a parallel encoder block (PEB) of the transformer and convolution to take their advantages in long-range dependency modeling and locality modeling for extracting semantics and preserving details. Second, we propose a novel random-connection attention (RCA) block, which has a reservoir structure to learn sparse attention via random connections during training. RCA encourages the target to attend to sparse relevant positions instead of all the large-area background pixels, resulting in more informative attention scores. It has fewer parameters and computations than the original self-attention in the transformer while performing better. Third, inspired by neural ordinary differential equations (ODE), we stack two PEBs with several residual connections as the basic encoder block to implement the Runge-Kutta method for solving ODE, which can effectively enhance the feature and suppress noise. Experiments on the public NUAA-SIRST dataset and IRSTD-1k dataset demonstrate the superiority of the RKformer over state-of-the-art methods.

Single frame infrared small target (SIRST) detection is useful for many practical applications, such as maritime rescue. However, SIRST detection is challenging due to the low-contrast between small targets and noisy background in infrared images. To address this challenge, we propose a novel FC3-Net by exploring feature compensation and cross-level correlation for SIRST detection. Specifically, FC3-Net consists of a Fine-detail guided Multi-level Feature Compensation (F-MFC) module, and a Cross-level Feature Correlation (CFC) module. The F-MFC module aims to compensate the information loss of details caused by the downsampling layers in convolutional neural networks (CNN) via aggregating features from multiple adjacent levels, so that the detail features of small targets can be propagated to the deeper layers of the network. Besides, to suppress the side impact of background noise, the CFC module constructs an energy filtering kernel based on the higher-level features with less background noise to filter out the noise in the middle-level features, and fuse them with the low-level ones to learn a strong target representation. Putting them together into the encoder-decoder structure, our FC3-Net could produce an accurate target mask with fine shape and details. Experiment results on the public NUAA-SIRST and IRSTD-1k datasets demonstrate that the proposed FC3-Net outperforms state-of-the-art methods in terms of both pixel-level and object-level metrics. The code will be released at https://github.com/IPIC-Lab/SIRST-Detection-FC3-Net.

Out-of-distribution (OOD) detection aims to identify OOD data based on representations extracted from well-trained deep models. However, existing methods largely ignore the reprogramming property of deep models and thus may not fully unleash their intrinsic strength: without modifying parameters of a well-trained deep model, we can reprogram this model for a new purpose via data-level manipulation (e.g., adding a specific feature perturbation). This property motivates us to reprogram a classification model to excel at OOD detection (a new task), and thus we propose a general methodology named watermarking in this paper. Specifically, we learn a unified pattern that is superimposed onto features of original data, and the model's detection capability is largely boosted after watermarking. Extensive experiments verify the effectiveness of watermarking, demonstrating the significance of the reprogramming property of deep models in OOD detection.

Although no specific domain knowledge is considered in the design, plain vision transformers have shown excellent performance in visual recognition tasks. However, little effort has been made to reveal the potential of such simple structures for pose estimation tasks. In this paper, we show the surprisingly good capabilities of plain vision transformers for pose estimation from various aspects, namely simplicity in model structure, scalability in model size, flexibility in training paradigm, and transferability of knowledge between models, through a simple baseline model called ViTPose. Specifically, ViTPose employs plain and non-hierarchical vision transformers as backbones to extract features for a given person instance and a lightweight decoder for pose estimation. It can be scaled up from 100M to 1B parameters by taking the advantages of the scalable model capacity and high parallelism of transformers, setting a new Pareto front between throughput and performance. Besides, ViTPose is very flexible regarding the attention type, input resolution, pre-training and finetuning strategy, as well as dealing with multiple pose tasks. We also empirically demonstrate that the knowledge of large ViTPose models can be easily transferred to small ones via a simple knowledge token. Experimental results show that our basic ViTPose model outperforms representative methods on the challenging MS COCO Keypoint Detection benchmark, while the largest model sets a new state-of-the-art. The code and models are available at https://github.com/ViTAE-Transformer/ViTPose.

Animal pose estimation and tracking (APT) is a fundamental task for detecting and tracking animal keypoints from a sequence of video frames. Previous animal-related datasets focus either on animal tracking or single-frame animal pose estimation, and never on both aspects. The lack of APT datasets hinders the development and evaluation of video-based animal pose estimation and tracking methods, limiting the applications in real world, e.g., understanding animal behavior in wildlife conservation. To fill this gap, we make the first step and propose APT-36K, i.e., the first large-scale benchmark for animal pose estimation and tracking. Specifically, APT-36K consists of 2,400 video clips collected and filtered from 30 animal species with 15 frames for each video, resulting in 36,000 frames in total. After manual annotation and careful double-check, high-quality keypoint and tracking annotations are provided for all the animal instances. Based on APT-36K, we benchmark several representative models on the following three tracks: (1) supervised animal pose estimation on a single frame under intra- and inter-domain transfer learning settings, (2) inter-species domain generalization test for unseen animals, and (3) animal pose estimation with animal tracking. Based on the experimental results, we gain some empirical insights and show that APT-36K provides a useful animal pose estimation and tracking benchmark, offering new challenges and opportunities for future research. The code and dataset will be made publicly available at https://github.com/pandorgan/APT-36K.

Perceptual organization is a challenging visual task that aims to perceive and group the individual visual element so that it is easy to understand the meaning of the scene as a whole. Most recent methods building upon advanced Convolutional Neural Network (CNN) come from learning discriminative representation and modeling context hierarchically. However, when the visual appearance difference between foreground and background is obscure, the performance of existing methods degrades significantly due to the visual ambiguity in the discrimination process. In this paper, we argue that the figure-ground assignment mechanism, which conforms to human vision cognitive theory, can be explored to empower CNN to achieve a robust perceptual organization despite visual ambiguity. Specifically, we present a novel Figure-Ground-Aided (FGA) module to learn the configural statistics of the visual scene and leverage it for the reduction of visual ambiguity. Particularly, we demonstrate the benefit of using stronger supervisory signals by teaching (FGA) module to perceive configural cues, \ie, convexity and lower region, that human deem important for the perceptual organization. Furthermore, an Interactive Enhancement Module (IEM) is devised to leverage such configural priors to assist representation learning, thereby achieving robust perception organization with complex visual ambiguities. In addition, a well-founded visual segregation test is designed to validate the capability of the proposed FGA mechanism explicitly. Comprehensive evaluation results demonstrate our proposed FGA mechanism can effectively enhance the capability of perception organization on various baseline models. Nevertheless, the model augmented via our proposed FGA mechanism also outperforms state-of-the-art approaches on four challenging real-world applications.

Recommender Systems (RS), as an efficient tool to discover users' interested items from a very large corpus, has attracted more and more attention from academia and industry. As the initial stage of RS, large-scale matching is fundamental yet challenging. A typical recipe is to learn user and item representations with a two-tower architecture and then calculate the similarity score between both representation vectors, which however still struggles in how to properly deal with negative samples. In this paper, we find that the common practice that randomly sampling negative samples from the entire space and treating them equally is not an optimal choice, since the negative samples from different sub-spaces at different stages have different importance to a matching model. To address this issue, we propose a novel method named Unbiased Model-Agnostic Matching Approach (UMA2). It consists of two basic modules including 1) General Matching Model (GMM), which is model-agnostic and can be implemented as any embedding-based two-tower models; and 2) Negative Samples Debias Network (NSDN), which discriminates negative samples by borrowing the idea of Inverse Propensity Weighting (IPW) and re-weighs the loss in GMM. UMA$^2$ seamlessly integrates these two modules in an end-to-end multi-task learning framework. Extensive experiments on both real-world offline dataset and online A/B test demonstrate its superiority over state-of-the-art methods.

Different from large-scale platforms such as Taobao and Amazon, CVR modeling in small-scale recommendation scenarios is more challenging due to the severe Data Distribution Fluctuation (DDF) issue. DDF prevents existing CVR models from being effective since 1) several months of data are needed to train CVR models sufficiently in small scenarios, leading to considerable distribution discrepancy between training and online serving; and 2) e-commerce promotions have significant impacts on small scenarios, leading to distribution uncertainty of the upcoming time period. In this work, we propose a novel CVR method named MetaCVR from a perspective of meta learning to address the DDF issue. Firstly, a base CVR model which consists of a Feature Representation Network (FRN) and output layers is designed and trained sufficiently with samples across months. Then we treat time periods with different data distributions as different occasions and obtain positive and negative prototypes for each occasion using the corresponding samples and the pre-trained FRN. Subsequently, a Distance Metric Network (DMN) is devised to calculate the distance metrics between each sample and all prototypes to facilitate mitigating the distribution uncertainty. At last, we develop an Ensemble Prediction Network (EPN) which incorporates the output of FRN and DMN to make the final CVR prediction. In this stage, we freeze the FRN and train the DMN and EPN with samples from recent time period, therefore effectively easing the distribution discrepancy. To the best of our knowledge, this is the first study of CVR prediction targeting the DDF issue in small-scale recommendation scenarios. Experimental results on real-world datasets validate the superiority of our MetaCVR and online A/B test also shows our model achieves impressive gains of 11.92% on PCVR and 8.64% on GMV.

In many classical e-commerce platforms, personalized recommendation has been proven to be of great business value, which can improve user satisfaction and increase the revenue of platforms. In this paper, we present a new recommendation problem, Trigger-Induced Recommendation (TIR), where users’ instant interest can be explicitly induced with a trigger item and follow-up related target items are recommended accordingly. TIR has become ubiquitous and popular in e-commerce platforms. In this paper, we figure out that although existing recommendation models are effective in traditional recommendation scenarios by mining users’ interests based on their massive historical behaviors, they are struggling in discovering users’ instant interests in the TIR scenario due to the discrepancy between these scenarios, resulting in inferior performance. To tackle the problem, we propose a novel recommendation method named Deep Interest Highlight Network (DIHN) for Click-Through Rate (CTR) prediction in TIR scenarios. It has three main components including 1) User Intent Network (UIN), which responds to generate a precise probability score to predict user’s intent on the trigger item; 2) Fusion Embedding Module (FEM), which adaptively fuses trigger item and target item embeddings based on the prediction from UIN; and (3) Hybrid Interest Extracting Module (HIEM), which can effectively highlight users’ instant interest from their behaviors based on the result of FEM. Extensive offline and online evaluations on a real-world e-commerce platform demonstrate the superiority of DIHN over state-of-the-art methods. Our code is available 1.

Abstract In this paper, we propose a unified information theoretic framework for learning-motivated methods aimed at odometry estimation, a crucial component of many robotics and vision tasks such as navigation and virtual reality where relative camera poses are required in real time. We formulate this problem as optimizing a variational information bottleneck objective function, which eliminates pose-irrelevant information from the latent representation. The proposed framework provides an elegant tool for performance evaluation and understanding in information-theoretic language. Specifically, we bound the generalization errors of the deep information bottleneck framework and the predictability of the latent representation. These provide not only a performance guarantee but also practical guidance for model design, sample collection, and sensor selection. Furthermore, the stochastic latent representation provides a natural uncertainty measure without the needs for extra structures or computations. Experiments on two well-known odometry datasets demonstrate the effectiveness of our method.

Although hand pose estimation has achieved a great success in recent years, there are still challenges with RGB-based estimation tasks, the most significant of which is the absence of labeled training data. At present, the synthetic dataset has plenty of images with accurate annotation, but the difference from real-world datasets affects generalization. Therefore, a transfer learning strategy, which tries to transfer knowledge from a labeled source domain to an unlabeled target domain, is a frequent solution. Existing methods such as mean-teacher, Cyclegan, and MCD will train models with the help of some easily accessible domains such as synthetic data. However, these methods are not guaranteed to operate well in real-world settings due to the domain shift. In this paper, we design a new unsupervised domain adaptation method named Multi-branch Adversarial Regressors (MarsDA) in hand pose estimation, where it could be better for feature migration. Specifically, we first generate pseudo-labels for unlabeled target domain data. Then, the new adversarial training loss between multiple regression branches we designed for hand pose estimation is introduced to narrow the domain gap. In this way, our model can reduce the noise of pseudo labels caused by the domain gap and improve the accuracy of pseudo labels. We evaluate our method on two publicly available real-world datasets, H3D and STB. Experimental results show that our method outperforms existing methods by a large margin.

Road segmentation from remote-sensing images is a challenging task with wide ranges of application potentials. Deep neural networks have advanced this field by leveraging the power of large-scale labeled data, which, however, are extremely expensive and time-consuming to acquire. One solution is to use cheap available data to train a model and deploy it to directly process the data from a specific application domain. Nevertheless, the well-known domain shift (DS) issue prevents the trained model from generalizing well on the target domain. In this article, we propose a novel stagewise domain adaptation model called RoadDA to address the DS issue in this field. In the first stage, RoadDA adapts the target domain features to align with the source ones via generative adversarial networks (GANs)-based interdomain adaptation. Specifically, a feature pyramid fusion module is devised to avoid information loss of long and thin roads and learn discriminative and robust features. Besides, to address the intradomain discrepancy in the target domain, in the second stage, we propose an adversarial self-training method. We generate the pseudo labels of the target domain using the trained generator and divide it to labeled easy split and unlabeled hard split based on the road confidence scores. The features of hard split are adapted to align with the easy ones using adversarial learning and the intradomain adaptation process is repeated to progressively improve the segmentation performance. Experiment results on two benchmarks demonstrate that RoadDA can efficiently reduce the domain gap and outperforms state-of-the-art methods. The code is available athttps://github.com/LANMNG/RoadDA.

Deep object detection models trained on clean images may not generalize well on degraded images due to the well-known domain shift issue. This hinders their application in real-life scenarios such as video surveillance and autonomous driving. Though domain adaptation methods can adapt the detection model from a labeled source domain to an unlabeled target domain, they struggle in dealing with open and compound degradation types. In this paper, we attempt to address this problem in the context of object detection by proposing a robust object Detector via Adversarial Novel Style Exploration (DANSE). Technically, DANSE first disentangles images into domain-irrelevant content representation and domain-specific style representation under an adversarial learning framework. Then, it explores the style space to discover diverse novel degradation styles that are complementary to those of the target domain images by leveraging a novelty regularizer and a diversity regularizer. The clean source domain images are transferred into these discovered styles by using a content-preserving regularizer to ensure realism. These transferred source domain images are combined with the target domain images and used to train a robust degradation-agnostic object detection model via adversarial domain adaptation. Experiments on both synthetic and real benchmark scenarios confirm the superiority of DANSE over state-of-the-art methods.

Previous deep learning-based video stabilizers require a large scale of paired unstable and stable videos for training, which are difficult to collect. Traditional trajectory-based stabilizers, on the other hand, divide the task into several sub-tasks and tackle them subsequently, which are fragile in textureless and occluded regions regarding the usage of hand-crafted features. In this paper, we attempt to tackle the video stabilization problem in a deep unsupervised learning manner, which borrows the divide-and-conquer idea from traditional stabilizers while leveraging the representation power of DNNs to handle the challenges in real-world scenarios. Technically, DUT is composed of a trajectory estimation stage and a trajectory smoothing stage. In the trajectory estimation stage, we first estimate the motion of keypoints, initialize and refine the motion of grids via a novel multi-homography estimation strategy and a motion refinement network, respectively, and get the grid-based trajectories via temporal association. In the trajectory smoothing stage, we devise a novel network to predict dynamic smoothing kernels for trajectory smoothing, which can well adapt to trajectories with different dynamic patterns. We exploit the spatial and temporal coherence of keypoints and grid vertices to formulate the training objectives, resulting in an unsupervised training scheme. Experiment results on public benchmarks show that DUT outperforms state-of-the-art methods both qualitatively and quantitatively. The source code is available at https://github.com/Annbless/DUTCode.

Prior human parsing models are limited to parsing humans into classes pre-defined in the training data, which is not flexible to generalize to unseen classes, e.g., new clothing in fashion analysis. In this paper, we propose a new problem named one-shot human parsing (OSHP) that requires to parse human into an open set of reference classes defined by any single reference example. During training, only base classes defined in the training set are exposed, which can overlap with part of reference classes. In this paper, we devise a novel Progressive One-shot Parsing network (POPNet) to address two critical challenges , i.e., testing bias and small sizes. POPNet consists of two collaborative metric learning modules named Attention Guidance Module and Nearest Centroid Module, which can learn representative prototypes for base classes and quickly transfer the ability to unseen classes during testing, thereby reducing testing bias. Moreover, POPNet adopts a progressive human parsing framework that can incorporate the learned knowledge of parent classes at the coarse granularity to help recognize the descendant classes at the fine granularity, thereby handling the small sizes issue. Experiments on the ATR-OS benchmark tailored for OSHP demonstrate POPNet outperforms other representative one-shot segmentation models by large margins and establishes a strong baseline. Source code can be found at https://github.com/Charleshhy/One-shot-Human-Parsing.

The travel marketing platform of Alibaba serves an indispensable role for hundreds of different travel scenarios from Fliggy, Taobao, Alipay apps, etc. To provide personalized recommendation service for users visiting different scenarios, there are two critical issues to be carefully addressed. First, since the traffic characteristics of different scenarios, e.g., individual data scale or representative topic, are significantly different, it is very challenging to train a unified model to serve all. Second, during the promotion period, the exposure of some specific items will be re-weighted due to manual intervention, resulting in biased logs, which will degrade the ranking model trained using these biased data. In this paper, we propose a novel Scenario-Aware Ranking Network (SAR-Net) to address these issues. SAR-Net harvests the abundant data from different scenarios by learning users' cross-scenario interests via two specific attention modules, which leverage the scenario features and item features to modulate the user behavior features, respectively. Then, taking the encoded features of previous module as input, a scenario-specific linear transformation layer is adopted to further extract scenario-specific features, followed by two groups of debias expert networks, i.e., scenario-specific experts and scenario-shared experts. They output intermediate results independently, which are further fused into the final result by a multi-scenario gating module. In addition, to mitigate the data fairness issue caused by manual intervention, we propose the concept of Fairness Coefficient (FC) to measures the importance of individual sample and use it to reweigh the prediction in the debias expert networks. Experiments on an offline dataset covering over 80 million users and 1.55 million travel items and an online A/B test demonstrate the effectiveness of our SAR-Net and its superiority over state-of-the-art methods. SAR-Net has also been deployed in the online travel marketing platform of Alibaba and is serving hundreds of travel scenarios.

Warping-based video stabilizers smooth camera trajectory by constraining each pixel's displacement and warp stabilized frames from unstable ones accordingly. However, since the view outside the boundary is not available during warping, the resulting holes around the boundary of the stabilized frame must be discarded (i.e., cropping) to maintain visual consistency, and thus does leads to a tradeoff between stability and cropping ratio. In this paper, we make a first attempt to address this issue by proposing a new Out-of-boundary View Synthesis (OVS) method. By the nature of spatial coherence between adjacent frames and within each frame, OVS extrapolates the out-of-boundary view by aligning adjacent frames to each reference one. Technically, it first calculates the optical flow and propagates it to the outer boundary region according to the affinity, and then warps pixels accordingly. OVS can be integrated into existing warping-based stabilizers as a plug-and-play module to significantly improve the cropping ratio of the stabilized results. In addition, stability is improved because the jitter amplification effect caused by cropping and resizing is reduced. Experimental results on the NUS benchmark show that OVS can improve the performance of five representative state-of-the-art methods in terms of objective metrics and subjective visual quality.1

CNN-based Spatiotemporal image fusion (STIF) methods have achieved better performance than traditional researches. However, most CNN-based methods fail to make full use of hierarchical features, and ignore the quality and distribution characteristics of feature maps in fine-grained STIF. In this paper, we propose a network with channel attention and feature super-resolution for STIF (CAFSRNet). First, our method uses the low resolution time-domain changing images as input to extract changes more accurately and simplify computational overhead. Second, channel attention mechanism is introduced into Cross-spatial Resolution Mapping module to make the network pay more attention to informative features. Third, by adding feature super-resolution into the supervision process, we enhance the distribution of feature maps and the quality of mapping results. The qualitative and quantitative experimental results on various datasets demonstrate the superiority of our proposed method over the state-of-the-art methods.

For hyperspectral classification, mixed pixels are usually the biggest reason to reduce the classification accuracy. To solve the problem, we apply the multi-label classification technique to the hyperspectral image classification. The approach Label-specific Features (LIFT) achieves state-of-the-art performance because the most distinctive features are constructed for each label. Clustering centers play an essential role in label-specific feature conversion. However, LIFT does not consider the relationship between positive and negative instances, and clustering centers on the training dataset is inconsistent with that of the original instances. In this paper, we propose a new algorithm called SMOTE_DFL, which can get better clustering centers through two strategies: 1) The spectral clustering algorithm SIA is introduced to focus on the local structure between positive and negative instances; 2) By oversampling training sample, the clustering center of training sample is close to that of the whole. Extensive experiments are conducted on three datasets. The results validate the superiority of SMOTE_DFL to other algorithms.

Convolutional neural network (CNN) is one of the most powerful tools to deal with computer vision tasks such as hyperspectral image (HSI) classification. While many studies using CNN focus on classification precision, few of them pay attention to the model size and running time. Some studies focus on lightweight neural networks for traditional RGB image processing tasks and achieve fantastic results, but none of them are designed for hyperspectral image processing. In this paper, a novel lightweight neural network designed for hyperspectral image classification is proposed to do fast HSI processing while maintaining high classification precision. The network uses the idea of feature reuse to reduce parameter size and improve convergence. Expansion convolution is adopted to overcome the defects which are brought by parameter reduction. The experiments show that the proposed network has SOTA level classification accuracy while maintaining high processing speed.

The presence of haze significantly reduces the quality of images. Researchers have designed a variety of algorithms for image dehazing (ID) to restore the quality of hazy images. However, there are few studies that summarize the deep learning (DL) based dehazing technologies. In this paper, we conduct a comprehensive survey on the recent proposed dehazing methods. Firstly, we conclude the commonly used datasets, loss functions and evaluation metrics. Secondly, we group the existing researches of ID into two major categories: supervised ID and unsupervised ID. The core ideas of various influential dehazing models are introduced. Finally, the open issues for future research on ID are pointed out.

Automatic image matting (AIM) refers to estimating the soft foreground from an arbitrary natural image without any auxiliary input like trimap, which is useful for image editing. Prior methods try to learn semantic features to aid the matting process while being limited to images with salient opaque foregrounds such as humans and animals. In this paper, we investigate the difficulties when extending them to natural images with salient transparent/meticulous foregrounds or non-salient foregrounds. To address the problem, a novel end-to-end matting network is proposed, which can predict a generalized trimap for any image of the above types as a unified semantic representation. Simultaneously, the learned semantic features guide the matting network to focus on the transition areas via an attention mechanism. We also construct a test set AIM-500 that contains 500 diverse natural images covering all types along with manually labeled alpha mattes, making it feasible to benchmark the generalization ability of AIM models. Results of the experiments demonstrate that our network trained on available composite matting datasets outperforms existing methods both objectively and subjectively. The source code and dataset are available at https://github.com/JizhiziLi/AIM.

Affordance detection refers to identifying the potential action possibilities of objects in an image, which is an important ability for robot perception and manipulation. To empower robots with this ability in unseen scenarios, we consider the challenging one-shot affordance detection problem in this paper, i.e., given a support image that depicts the action purpose, all objects in a scene with the common affordance should be detected. To this end, we devise a One-Shot Affordance Detection (OS-AD) network that firstly estimates the purpose and then transfers it to help detect the common affordance from all candidate images. Through collaboration learning, OS-AD can capture the common characteristics between objects having the same underlying affordance and learn a good adaptation capability for perceiving unseen affordances. Besides, we build a Purpose-driven Affordance Dataset (PAD) by collecting and labeling 4k images from 31 affordance and 72 object categories. Experimental results demonstrate the superiority of our model over previous representative ones in terms of both objective metrics and visual quality. The benchmark suite is at ProjectPage.

Unsupervised domain adaptation (UDA) for semantic segmentation aims to adapt a segmentation model trained on the labeled source domain to the unlabeled target domain. Existing methods try to learn domain invariant features while suffering from large domain gaps that make it difficult to correctly align discrepant features, especially in the initial training phase. To address this issue, we propose a novel Dual Soft-Paste (DSP) method in this paper. Specifically, DSP selects some classes from a source domain image using a long-tail class first sampling strategy and softly pastes the corresponding image patch on both the source and target training images with a fusion weight. Technically, we adopt the mean teacher framework for domain adaptation, where the pasted source and target images go through the student network while the original target image goes through the teacher network. Output-level alignment is carried out by aligning the probability maps of the target fused image from both networks using a weighted cross-entropy loss. In addition, feature-level alignment is carried out by aligning the feature maps of the source and target images from student network using a weighted maximum mean discrepancy loss. DSP facilitates the model learning domain-invariant features from the intermediate domains, leading to faster convergence and better performance. Experiments on two challenging benchmarks demonstrate the superiority of DSP over state-of-the-art methods. Code is available at https://github.com/GaoLii/DSP.

Recently, there has been an increasing concern about the privacy issue raised by using personally identifiable information in machine learning. However, previous portrait matting methods were all based on identifiable portrait images. To fill the gap, we present P3M-10k in this paper, which is the first large-scale anonymized benchmark for Privacy-Preserving Portrait Matting. P3M-10k consists of 10,000 high-resolution face-blurred portrait images along with high-quality alpha mattes. We systematically evaluate both trimap-free and trimap-based matting methods on P3M-10k and find that existing matting methods show different generalization capabilities when following the Privacy-Preserving Training (PPT) setting, i.e., training on face-blurred images and testing on arbitrary images. To devise a better trimap-free portrait matting model, we propose P3M-Net, which leverages the power of a unified framework for both semantic perception and detail matting, and specifically emphasizes the interaction between them and the encoder to facilitate the matting process. Extensive experiments on P3M-10k demonstrate that P3M-Net outperforms the state-of-the-art methods in terms of both objective metrics and subjective visual quality. Besides, it shows good generalization capacity under the PPT setting, confirming the value of P3M-10k for facilitating future research and enabling potential real-world applications. The source code and dataset are available at https://github.com/JizhiziLi/P3M.

Detection transformers have recently shown promising object detection results and attracted increasing attention. However, how to develop effective domain adaptation techniques to improve its cross-domain performance remains unexplored and unclear. In this paper, we delve into this topic and empirically find that direct feature distribution alignment on the CNN backbone only brings limited improvements, as it does not guarantee domain-invariant sequence features in the transformer for prediction. To address this issue, we propose a novel Sequence Feature Alignment (SFA) method that is specially designed for the adaptation of detection transformers. Technically, SFA consists of a domain query-based feature alignment (DQFA) module and a token-wise feature alignment (TDA) module. In DQFA, a novel domain query is used to aggregate and align global context from the token sequence of both domains. DQFA reduces the domain discrepancy in global feature representations and object relations when deploying in the transformer encoder and decoder, respectively. Meanwhile, TDA aligns token features in the sequence from both domains, which reduces the domain gaps in local and instance-level feature representations in the transformer encoder and decoder, respectively. Besides, a novel bipartite matching consistency loss is proposed to enhance the feature discriminability for robust object detection. Experiments on three challenging benchmarks show that SFA outperforms state-of-the-art domain adaptive object detection methods. Code has been made available at: https://github.com/encounter1997/SFA.

Transformers have shown great potential in various computer vision tasks owing to their strong capability in modeling long-range dependency using the self-attention mechanism. Nevertheless, vision transformers treat an image as 1D sequence of visual tokens, lacking an intrinsic inductive bias (IB) in modeling local visual structures and dealing with scale variance. Alternatively, they require large-scale training data and longer training schedules to learn the IB implicitly. In this paper, we propose a new Vision Transformer Advanced by Exploring intrinsic IB from convolutions, i.e., ViTAE. Technically, ViTAE has several spatial pyramid reduction modules to downsample and embed the input image into tokens with rich multi-scale context by using multiple convolutions with different dilation rates. In this way, it acquires an intrinsic scale invariance IB and is able to learn robust feature representation for objects at various scales. Moreover, in each transformer layer, ViTAE has a convolution block in parallel to the multi-head self-attention module, whose features are fused and fed into the feed-forward network. Consequently, it has the intrinsic locality IB and is able to learn local features and global dependencies collaboratively. Experiments on ImageNet as well as downstream tasks prove the superiority of ViTAE over the baseline transformer and concurrent works. Source code and pretrained models will be available at https://github.com/Annbless/ViTAE.

Conversion Rate (CVR) prediction in modern industrial e-commerce platforms is becoming increasingly important, which directly contributes to the final revenue. In order to address the well-known sample selection bias (SSB) and data sparsity (DS) issues encountered during CVR modeling, the abundant labeled macro behaviors (i.e., user's interactions with items) are used. Nonetheless, we observe that several purchase-related micro behaviors (i.e., user's interactions with specific components on the item detail page) can supplement fine-grained cues for CVR prediction. Motivated by this observation, we propose a novel CVR prediction method by Hierarchically Modeling both Micro and Macro behaviors (HM3). Specifically, we first construct a complete user sequential behavior graph to hierarchically represent micro behaviors and macro behaviors as one-hop and two-hop post-click nodes. Then, we embody HM3 as a multi-head deep neural network, which predicts six probability variables corresponding to explicit sub-paths in the graph. They are further combined into the prediction targets of four auxiliary tasks as well as the final CVR according to the conditional probability rule defined on the graph. By employing multi-task learning and leveraging the abundant supervisory labels from micro and macro behaviors, HM3 can be trained end-to-end and address the SSB and DS issues. Extensive experiments on both offline and online settings demonstrate the superiority of the proposed HM3 over representative state-of-the-art methods.

Abstract Chest X‐ray images taken under different conditions follow different distributions, preventing the models trained on a domain from generalising well on the other domain. In this paper, a generative domain adaptation (GDA) method is proposed to address this issue and facilitate the learning process for downstream analysis. GDA adapts different domains to a virtual common one where images are aligned at the appearance level. To this end, a domain shared generator is used to transform the input images and two competitive discriminators are used to adversarially supervise the transforming process. The domain discriminator drives the generator to narrow the domain gap while the fidelity discriminator forces the generator to keep the inherent information. Moreover, a specific classification or detection network is attached to the generator to supervise it in a task‐oriented manner. Experiment results on a large‐scale dataset containing 46k chest X‐ray images demonstrate that GDA outperforms representative domain adaptation methods by a large margin for both disease classification and lesion detection as well as provides useful transformed images to assist experts for diagnosis.

Digital twin (DT) systems that replicate the physical world digitally are powerful tools for monitoring physical systems and evaluating algorithms, but current DT systems are commonly not applicable for robotic deployment and investigation. Meanwhile, current 3-D simulation-based robotic platforms do not model the dynamics of the physical world on-the-fly as done in DT systems, limiting their potential for the development of robotics in challenging environments. To tackle this issue, we propose the first robot-centered smart DT framework, namely, Terra, to facilitate the deployment of robots in challenging environments. The proposed Terra framework introduces a comprehensive DT representation to encode the useful real-time dynamics of both the physical world and the robot agent deployed therein. A multiview multimodality perception module is further devised for Terra to obtain high-level semantics and deliver a precise description of the current status of the environment and the robot agent. By mapping the perceived results to the virtual replica of the physical environment, Terra actively updates the action policy and sends it back to the agent, forming an integral and real-time information feedback loop. In practice, to help demonstrate the effectiveness and feasibility of the proposed framework, we deliberately set up a challenging unordered physical environment with many obstacles and a very simple robot aiming to fulfill a navigation task. Empirical results show that the proposed Terra framework successfully facilitates the robot to accomplish the task without causing hazards.

In the Internet-of-Things (IoT) era, billions of sensors and devices collect and process data from the environment, transmit them to cloud centers, and receive feedback via the Internet for connectivity and perception. However, transmitting massive amounts of heterogeneous data, perceiving complex environments from these data, and then making smart decisions in a timely manner are difficult. Artificial intelligence (AI), especially deep learning, is now a proven success in various areas, including computer vision, speech recognition, and natural language processing. AI introduced into the IoT heralds the era of AI of things (AIoT). This article presents a comprehensive survey on AIoT to show how AI can empower the IoT to make it faster, smarter, greener, and safer. Specifically, we briefly present the AIoT architecture in the context of cloud computing, fog computing, and edge computing. Then, we present progress in AI research for IoT from four perspectives: 1) perceiving; 2) learning; 3) reasoning; and 4) behaving. Next, we summarize some promising applications of AIoT that are likely to profoundly reshape our world. Finally, we highlight the challenges facing AIoT and some potential research opportunities.

Underwater imaging often suffers from color cast and contrast degradation due to range-dependent medium absorption and light scattering. Introducing image statistics as prior has been proved to be an effective solution for underwater image enhancement. However, relative to the modal divergence of light propagation and underwater scenery, the existing methods are limited in representing the inherent statistics of underwater images resulting in color artifacts and haze residuals. To address this problem, this article proposes a convolutional neural network (CNN)-based framework to learn hierarchical statistical features related to color cast and contrast degradation and to leverage them for underwater image enhancement. Specifically, a pixel disruption strategy is first proposed to suppress intrinsic colors’ influence and facilitate modeling a unified statistical representation of underwater image. Then, considering the local variation of depth of field, two parallel sub-networks: Color Correction Network (CC-Net) and Contrast Enhancement Network (CE-Net) are presented. The CC-Net and CE-Net can generate pixel-wise color cast and transmission map and achieve spatial-varied color correction and contrast enhancement. Moreover, to address the issue of insufficient training data, an imaging model-based synthesis method that incorporates pixel disruption strategy is presented to generate underwater patches with global degradation consistency. Quantitative and subjective evaluations demonstrate that our proposed method achieves state-of-the-art performance.

Human parsing, or human body part semantic segmentation, has been an active research topic due to its wide potential applications. In this paper, we propose a novel GRAph PYramid Mutual Learning (Grapy-ML) method to address the cross-dataset human parsing problem, where the annotations are at different granularities. Starting from the prior knowledge of the human body hierarchical structure, we devise a graph pyramid module (GPM) by stacking three levels of graph structures from coarse granularity to fine granularity subsequently. At each level, GPM utilizes the self-attention mechanism to model the correlations between context nodes. Then, it adopts a top-down mechanism to progressively refine the hierarchical features through all the levels. GPM also enables efficient mutual learning. Specifically, the network weights of the first two levels are shared to exchange the learned coarse-granularity information across different datasets. By making use of the multi-granularity labels, Grapy-ML learns a more discriminative feature representation and achieves state-of-the-art performance, which is demonstrated by extensive experiments on the three popular benchmarks, e.g. CIHP dataset. The source code is publicly available at https://github.com/Charleshhy/Grapy-ML.

State-of-the-art image classification algorithms building upon convolutional neural networks (CNNs) are commonly trained on large annotated datasets of high-quality images. When applied to low-quality images, they will suffer a significant degradation in performance, since the structural and statistical properties of pixels in the neighborhood are obstructed by image degradation. To address this problem, this paper proposes a novel deep degradation prior for low-quality image classification. It is based on statistical observations that, in the deep representation space, image patches with structural similarity have uniform distribution even if they come from different images, and the distributions of corresponding patches in low- and high-quality images have uniform margins under the same degradation condition. Therefore, we propose a feature de-drifting module (FDM) to learn the mapping relationship between deep representations of low- and high- quality images, and leverage it as a deep degradation prior (DDP) for low-quality image classification. Since the statistical properties are independent to image content, deep degradation prior can be learned on a training set of limited images without supervision of semantic labels and served in a form of “plugging-in” module of the existing classification networks to improve their performance on degraded images. Evaluations on the benchmark dataset ImageNet-C demonstrate that our proposed DDP can improve the accuracy of the pre-trained network model by more than 20% under various degradation conditions. Even under the extreme setting that only 10 images from CUB-C dataset are used for the training of DDP, our method improves the accuracy of VGG16 on ImageNet-C from 37% to 55%.

Existing deep learning-based pan-sharpening methods mainly learn spatial information from a high-resolution (HR) panchromatic (PAN) image for each spectral channel. However, due to the own characteristics of remote sensing image data, the spatial information of PAN image often shows weak correlation with some spectral channel, especially for channels non-overlapped by PAN channel. In this paper, we propose a parallel pyramid network (PPN) for pan-sharpening. First, a three-branch parallel structure is proposed for dealing with PAN image detail, multispectral (MS) images detail and spectral property respectively. Second, pyramid network structure is introduced in two detail branches to solve the problem of weak correlation due to scale difference. Third, the feature level fusion in two detail branches is implemented, which utilizes redundancy between channels to solve detail representation of channels non-overlapped by PAN channel. The qualitative and quantitative experimental results on various data sets demonstrate the superiority of our proposed method over the state-of-the-art methods.

Increasing the visibility of nighttime hazy images is challenging because of uneven illumination from active artificial light sources and haze absorbing/scattering. The absence of large-scale benchmark datasets hampers progress in this area. To address this issue, we propose a novel synthetic method called 3R to simulate nighttime hazy images from daytime clear images, which first reconstructs the scene geometry, then simulates the light rays and object reflectance, and finally renders the haze effects. Based on it, we generate realistic nighttime hazy images by sampling real-world light colors from a prior empirical distribution. Experiments on the synthetic benchmark show that the degrading factors jointly reduce the image quality. To address this issue, we propose an optimal-scale maximum reflectance prior to disentangle the color correction from haze removal and address them sequentially. Besides, we also devise a simple but effective learning-based baseline which has an encoder-decoder structure based on the MobileNet-v2 backbone. Experiment results demonstrate their superiority over state-of-the-art methods in terms of both image quality and runtime. Both the dataset and source code will be available at https://github.com/chaimi2013/3R.

Attention modules have been demonstrated effective in strengthening the representation ability of a neural network via reweighting spatial or channel features or stacking both operations sequentially. However, designing the structures of different attention operations requires a bulk of computation and extensive expertise. In this paper, we devise an Auto Learning Attention (AutoLA) method, which is the first attempt on automatic attention design. Specifically, we define a novel attention module named high order group attention (HOGA) as a directed acyclic graph (DAG) where each group represents a node, and each edge represents an operation of heterogeneous attentions. A typical HOGA architecture can be searched automatically via the differential AutoLA method within 1 GPU day using the ResNet-20 backbone on CIFAR10. Further, the searched attention module can generalize to various backbones as a plug-and-play component and outperforms popular manually designed channel and spatial attentions for many vision tasks, including image classification on CIFAR100 and ImageNet, object detection and human keypoint detection on COCO dataset. The code will be released.

Recommender system, as an essential part of modern e-commerce, consists of two fundamental modules, namely Click-Through Rate (CTR) and Conversion Rate (CVR) prediction. While CVR has a direct impact on the purchasing volume, its prediction is well-known challenging due to the Sample Selection Bias (SSB) and Data Sparsity (DS) issues. Although existing methods, typically built on the user sequential behavior path "impression->click->purchase", is effective for dealing with SSB issue, they still struggle to address the DS issue due to rare purchase training samples. Observing that users always take several purchase-related actions after clicking, we propose a novel idea of post-click behavior decomposition. Specifically, disjoint purchase-related Deterministic Action (DAction) and Other Action (OAction) are inserted between click and purchase in parallel, forming a novel user sequential behavior graph "impression->click->D(O)Action->purchase". Defining model on this graph enables to leverage all the impression samples over the entire space and extra abundant supervised signals from D(O)Action, which will effectively address the SSB and DS issues together. To this end, we devise a novel deep recommendation model named Elaborated Entire Space Supervised Multi-task Model (ESM2). According to the conditional probability rule defined on the graph, it employs multi-task learning to predict some decomposed sub-targets in parallel and compose them sequentially to formulate the final CVR. Extensive experiments on both offline and online environments demonstrate the superiority of ESM2 over state-of-the-art models. The source code and dataset will be released.

Natural scene text usually carries considerable abstract semantic information, which is closely related to the surrounding environment. Thus, natural scene text detection plays a vital role in image content retrieval and understanding. In this study, the authors propose a novel specific category region proposal network (SCRPN) based on maximally stable extremal regions (MSER) and fully convolutional network (FCN) for natural scene text detection. First, FCN for pixel‐level recognition is utilised to obtain the text saliency map and MSER is used to obtain oversegmented regions. Then, the multiple features of oversegmented regions and text saliency map are used for region aggregation. Next, single‐linkage clustering method is adopted to cluster the segmentation regions to obtain a hierarchical structure of text region proposals. Finally, for the top‐ranking region proposals, SCRPN built an end‐to‐end pipeline for scene text detection directly. Experiments on street view text and international conference on document analysis and recognition (ICDAR) 2013 have demonstrated the effectiveness of SCRPN for generating the text proposals. SCRPN could work with various two‐stage text detection networks; thus, faster region convolutional neural network was used as the text detection framework to evaluate the performance of SCRPN in the ICDAR 2015 and MSRA‐TD500 benchmarks. The experimental results confirmed that SCRPN makes text detection more robust in complex scenarios.

Single image dehazing is a critical image pre-processing step for subsequent high-level computer vision tasks. However, it remains challenging due to its ill-posed nature. Existing dehazing models tend to suffer from model overcomplexity and computational inefficiency or have limited representation capacity. To tackle these challenges, here, we propose a fast and accurate multi-scale end-to-end dehazing network, called FAMED-Net, which comprises encoders at three scales and a fusion module to efficiently and directly learn the haze-free image. Each encoder consists of cascaded and densely connected point-wise convolutional layers and pooling layers. Since no larger convolutional kernels are used and features are reused layer-by-layer, FAMED-Net is lightweight and computationally efficient. Thorough empirical studies on public synthetic datasets (including RESIDE) and real-world hazy images demonstrate the superiority of FAMED-Net over other representative state-of-the-art models with respect to model complexity, computational efficiency, restoration accuracy, and cross-set generalization. The code will be made publicly available.

Developing effective and efficient recommendation methods is very challenging for modern e-commerce platforms. Generally speaking, two essential modules named “ClickThrough Rate Prediction” (CTR) and “Conversion Rate Prediction” (CVR) are included, where CVR module is a crucial factor that affects the final purchasing volume directly. However, it is indeed very challenging due to its sparseness nature. In this paper, we tackle this problem by proposing multiLevel Deep Cascade Trees (ldcTree), which is a novel decision tree ensemble approach. It leverages deep cascade structures by stacking Gradient Boosting Decision Trees (GBDT) to effectively learn feature representation. In addition, we propose to utilize the cross-entropy in each tree of the preceding GBDT as the input feature representation for next level GBDT, which has a clear explanation, i.e., a traversal from root to leaf nodes in the next level GBDT corresponds to the combination of certain traversals in the preceding GBDT. The deep cascade structure and the combination rule enable the proposed ldcTree to have a stronger distributed feature representation ability. Moreover, inspired by ensemble learning, we propose an Ensemble ldcTree (E-ldcTree) to encourage the model’s diversity and enhance the representation ability further. Finally, we propose an improved Feature learning method based on EldcTree (F-EldcTree) for taking adequate use of weak and strong correlation features identified by pretrained GBDT models. Experimental results on off-line data set and online deployment demonstrate the effectiveness of the proposed methods.

Generating an image from a given text description has two goals: visual realism and semantic consistency. Although significant progress has been made in generating high-quality and visually realistic images using generative adversarial networks, guaranteeing semantic consistency between the text description and visual content remains very challenging. In this paper, we address this problem by proposing a novel global-local attentive and semantic-preserving text-to-image-to-text framework called MirrorGAN. MirrorGAN exploits the idea of learning text-to-image generation by redescription and consists of three modules: a semantic text embedding module (STEM), a global-local collaborative attentive module for cascaded image generation (GLAM), and a semantic text regeneration and alignment module (STREAM). STEM generates word- and sentence-level embeddings. GLAM has a cascaded architecture for generating target images from coarse to fine scales, leveraging both local word attention and global sentence attention to progressively enhance the diversity and semantic consistency of the generated images. STREAM seeks to regenerate the text description from the generated image, which semantically aligns with the given text description. Thorough experiments on two public benchmark datasets demonstrate the superiority of MirrorGAN over other representative state-of-the-art methods.

Texture recognition is a challenging visual task as multiple perceptual attributes may be perceived from the same texture image when combined with different spatial context. Some recent works building upon Convolutional Neural Network (CNN) incorporate feature encoding with orderless aggregating to provide invariance to spatial layouts. However, these existing methods ignore visual texture attributes, which are important cues for describing the real-world texture images, resulting in incomplete description and inaccurate recognition. To address this problem, we propose a novel deep Multiple-Attribute-Perceived Network (MAP-Net) by progressively learning visual texture attributes in a mutually reinforced manner. Specifically, a multi-branch network architecture is devised, in which cascaded global contexts are learned by introducing similarity constraint at each branch, and leveraged as guidance of spatial feature encoding at next branch through an attribute transfer scheme. To enhance the modeling capability of spatial transformation, a deformable pooling strategy is introduced to augment the spatial sampling with adaptive offsets to the global context, leading to perceive new visual attributes. An attribute fusion module is then introduced to jointly utilize the perceived visual attributes and the abstracted semantic concepts at each branch. Experimental results on the five most challenging texture recognition datasets have demonstrated the superiority of the proposed model against the state-of-the-arts.

Contrast enhancement and noise removal are coupled problems for low-light image enhancement. The existing Retinex based methods do not take the coupling relation into consideration, resulting in under or over-smoothing of the enhanced images. To address this issue, this paper presents a novel progressive Retinex framework, in which illumination and noise of low-light image are perceived in a mutually reinforced manner, leading to noise reduction low-light enhancement results. Specifically, two fully pointwise convolutional neural networks are devised to model the statistical regularities of ambient light and image noise respectively, and to leverage them as constraints to facilitate the mutual learning process. The proposed method not only suppresses the interference caused by the ambiguity between tiny textures and image noises, but also greatly improves the computational efficiency. Moreover, to solve the problem of insufficient training data, we propose an image synthesis strategy based on camera imaging model, which generates color images corrupted by illumination-dependent noises. Experimental results on both synthetic and real low-light images demonstrate the superiority of our proposed approaches against the State-Of-The-Art (SOTA) low-light enhancement methods.

Text-to-image generation, i.e. generating an image given a text description, is a very challenging task due to the significant semantic gap between the two domains. Humans, however, tackle this problem intelligently. We learn from diverse objects to form a solid prior about semantics, textures, colors, shapes, and layouts. Given a text description, we immediately imagine an overall visual impression using this prior and, based on this, we draw a picture by progressively adding more and more details. In this paper, and inspired by this process, we propose a novel text-to-image method called LeicaGAN to combine the above three phases in a unified framework. First, we formulate the multiple priors learning phase as a textual-visual co-embedding (TVE) comprising a text-image encoder for learning semantic, texture, and color priors and a text-mask encoder for learning shape and layout priors. Then, we formulate the imagination phase as multiple priors aggregation (MPA) by combining these complementary priors and adding noise for diversity. Lastly, we formulate the creation phase by using a cascaded attentive generator (CAG) to progressively draw a picture from coarse to fine. We leverage adversarial learning for LeicaGAN to enforce semantic consistency and visual realism. Thorough experiments on two public benchmark datasets demonstrate LeicaGAN's superiority over the baseline method. Code has been made available at https://github.com/qiaott/LeicaGAN.

Unsupervised domain adaptation (UDA) aims to enhance the generalization capability of a certain model from a source domain to a target domain. UDA is of particular significance since no extra effort is devoted to annotating target domain samples. However, the different data distributions in the two domains, or \emph{domain shift/discrepancy}, inevitably compromise the UDA performance. Although there has been a progress in matching the marginal distributions between two domains, the classifier favors the source domain features and makes incorrect predictions on the target domain due to category-agnostic feature alignment. In this paper, we propose a novel category anchor-guided (CAG) UDA model for semantic segmentation, which explicitly enforces category-aware feature alignment to learn shared discriminative features and classifiers simultaneously. First, the category-wise centroids of the source domain features are used as guided anchors to identify the active features in the target domain and also assign them pseudo-labels. Then, we leverage an anchor-based pixel-level distance loss and a discriminative loss to drive the intra-category features closer and the inter-category features further apart, respectively. Finally, we devise a stagewise training mechanism to reduce the error accumulation and adapt the proposed model progressively. Experiments on both the GTA5$\rightarrow $Cityscapes and SYNTHIA$\rightarrow $Cityscapes scenarios demonstrate the superiority of our CAG-UDA model over the state-of-the-art methods. The code is available at \url{https://github.com/RogerZhangzz/CAG\_UDA}.

Modeling statistical regularity plays an essential role in ill-posed image processing problems. Recently, deep learning based methods have been presented to implicitly learn statistical representation of pixel distributions in natural images and leverage it as a constraint to facilitate subsequent tasks, such as color constancy and image dehazing. However, the existing CNN architecture is prone to variability and diversity of pixel intensity within and between local regions, which may result in inaccurate statistical representation. To address this problem, this paper presents a novel fully point-wise CNN architecture for modeling statistical regularities in natural images. Specifically, we propose to randomly shuffle the pixels in the origin images and leverage the shuffled image as input to make CNN more concerned with the statistical properties. Moreover, since the pixels in the shuffled image are independent identically distributed, we can replace all the large convolution kernels in CNN with point-wise (1*1) convolution kernels while maintaining the representation ability. Experimental results on two applications: color constancy and image dehazing, demonstrate the superiority of our proposed network over the existing architectures, i.e., using 1/10~1/100 network parameters and computational cost while achieving comparable performance.

In this paper, we address a haze removal problem from a single nighttime image, even in the presence of varicolored and non-uniform illumination. The core idea lies in a novel maximum reflectance prior. We first introduce the nighttime hazy imaging model, which includes a local ambient illumination item in both direct attenuation term and scattering term. Then, we propose a simple but effective image prior, maximum reflectance prior, to estimate the varying ambient illumination. The maximum reflectance prior is based on a key observation: for most daytime haze-free image patches, each color channel has very high intensity at some pixels. For the nighttime haze image, the local maximum intensities at each color channel are mainly contributed by the ambient illumination. Therefore, we can directly estimate the ambient illumination and transmission map, and consequently restore a high quality haze-free image. Experimental results on various nighttime hazy images demonstrate the effectiveness of the proposed approach. In particular, our approach has the advantage of computational efficiency, which is 10-100 times faster than state-of-the-art methods.

Underwater images often suffer from color distortion and visibility degradation due to the light absorption and scattering. Existing methods utilize various assumptions/constrains to achieve reasonable solutions for underwater image enhancement. However, these methods share the common limitation that the adopted assumptions may not work for some particular scenes. To address this problem, this paper proposes an end to end framework for underwater image enhancement, where a CNN-based network called UIE-Net is presented. The UIE-net is trained with two tasks, color correction and haze removal. This unified training approach enables learning a strong feature representation for both tasks simultaneously. For better extracting the inherent features in local patches, a pixels disrupting strategy is exploited in the proposed learning framework, which significantly improves the convergent speed and accuracy. To handle the training of UIE-net, we synthesize 200000 training images based on the physical underwater imaging model. Experiments on benchmark underwater images for cross-scenes show that UIE-net achieves superior performance over existing methods.

Depth maps captured by RGB-D cameras are often noisy and incomplete at edge regions. Most existing methods assume that there is a co-occurrence of edges in depth map and its corresponding color image, and improve the quality of depth map guided by the color image. However, when the color image is noisy or richly detailed, the high frequency artifacts will be introduced into depth map. In this paper, we propose a deep residual network based on deep fusion and local linear regularization for guided depth enhancement. The presented scheme can effectively extract the correlation between depth map and color image in the deep feature space. To reduce the difficulty of training, a specific layer of network which introduces a local linear regularization constraint on the output depth is designed. Experiments on various applications, including depth denoising, super-resolution and inpainting, demonstrate the effectiveness and reliability of our proposed approach.

Light field cameras have been recently shown to be very effective in applications such as multifocusing and 3D reconstruction. These cameras can provide depth cues from both defocus and correspondence in a single snapshot. In this paper, we present a fast response aggregation framework for depth estimation by jointly using defocus and correspondence cues. Different from existing approaches, we perform a fast gradient preserving filtering in a label domain, instead of in a depth domain, to efficiently compute a dense depth map. The proposed approach comprises of three steps: 1) constructing defocus and correspondence response volumes, 2) adaptively smoothing the two volumes and performing Winner-Takes-All label selections, and 3) post-processing by using nonlocal image guided averaging. With such a compact framework, currently best depth estimation results can be achieved. This compact framework is suitable for various applications such as object segmentation and surface reconstruction.

Reconstructing a full-resolution stereoscopic video from an asymmetric stereoscopic video is a challenging task. The existing approaches require depth information, which imposes an additional challenge in data acquisition. In this paper, we propose a novel scheme that is capable of obtaining super-resolution and depth estimation simultaneously from an asymmetric stereoscopic video. The proposed scheme models the video super-resolution and stereo matching with a unified energy function. Then, we apply an alternating optimization method to minimize this energy function, which can be implemented with a two-step algorithm. In the first step we calculate the initial depth map by using a region-based cooperative optimization technique while considering the temporal consistency in video. In the second step we resolve the super-resolution problem under the guidance of the depth information. It is effective because each step benefits from the additional improvement over the previous step. We iteratively update the two steps until stable depth and super-resolution results are obtained. We have conducted a series of experiments on public stereoscopic video sequences to evaluate the performance of the proposed method. Both objective indexes and subjective visual comparisons verify that the proposed scheme can achieve satisfactory super-resolution results and high-quality depth map simultaneously. In particular, the subjective evaluation experiments on a 3-D monitor show that this scheme outperforms others and achieves the best visual sharpness.

Image guided filtering has been widely used in many image processing applications. However, it is a local filtering method and has limited propagation ability. In this paper, we propose a new image filtering method: nonlocal image guided averaging (NLGA). Derived from a nonlocal linear model, the proposed method can utilize the nonlocal similarity of the guidance image, so that it can propagate nonlocal information reliably. Consequently, NLGA can obtain a sharper filtering results in the edge regions and more smooth results in the smooth regions. It shows superiority over image guided filtering in different applications, such as image dehazing, depth map super-resolution and image denoising.

Nighttime haze removal is important for different applications such as nighttime video surveillance in haze environment. Different from the imaging conditions in the daytime, nighttime haze images may suffer from non-uniform illumination from artificial light sources. In this paper, we proposed a novel efficient dehazing method with illumination estimation for nighttime haze condition. First, we estimate the light intensity and enhance it to obtain an illumination balanced result. Then, we process a color correction step after estimating the color characteristics of the incident light. Finally, we remove the haze by using the dark channel prior along with estimating the pointwise environmental light. Experimental results show that the proposed method can achieve both illumination balanced and haze free results. Moreover, it also has good color rendition ability.

Stereo image applications are becoming more and more prevalent. However, there has been little research on stereo image enhancement. In this paper, we address the challenging problem of underwater stereo image enhancement. A new underwater imaging model is proposed and it can better describe the degradation of underwater images including color distortion and contrast attenuation. In addition, a novel observation that the intensity of the water part within the image is mainly contributed by the scattering light is also proposed. Coupling the proposed model and prior together, the parameters of scattering light can be estimated. Then an iterative approach to process stereo matching and stereo image enhancement alternatively is presented, which can significantly improve the quality of the images and depth maps. The experimental results demonstrate that the proposed method can significantly enhance the image visibility and achieve better depth perception.

Detecting moving objects from background in video sequences is the first step of many image applications. The background can be divided into two types according to whether the pixel values of it are variable or not: static one and dynamic one. How to correctly detect moving foreground objects from dynamic scenes is a difficult problem because of the similarity between the moving foreground and the variable background. In this study, a new method for non‐parametric background segmentation of dynamic scenes is proposed. Here the background is described by two interrelated models. One of them is called the self‐model, which concerns with the recently observed pixel values at the same position, and the other one is called the neighbourhood‐model, which is described by the pixel values of the neighbourhood. The author's method can accurately detect the dynamic background. To correctly detect pixels in the foreground as much as possible, the authors also propose an adaptive threshold for foreground decision based on the background characteristics. All of the above detection processes can be done in real time. Experimental results on public dataset demonstrate that the proposed method outperforms the state‐of‐the‐art for background segmentation in dynamic scenes.

Mixed-resolution approach serves as a feasible solution to 3D video data reduction in limited bandwidth network environments, i.e., mobile networks. In this paper, we propose a simultaneous method for video super-resolution and high-quality depth estimation of mixed-resolution 3D video. Our method tackles the problem in a joint manner: i). Depth Estimation. We calculate the initial depths by stereo matching, and then warp them according to the optical flow field. ii). Video Super-resolution. Under the guidance of the warped depth information, we resolve the super-resolution problem by a fusion method, which involves a mapping step and a nonlocal reconstruction step. We run the above two steps iteratively until obtaining the stable depth and super-resolution result. The experimental results on public 3D video sequences verify the effectiveness of our proposed method.