Layer pruning is a viable technique for compressing large language models while achieving acceleration proportional to the pruning ratio. In this work, we identify that removing any layer induces a magnitude gap in hidden states, and demonstrate that a simple compensation operation leads to superior performance in iterative layer pruning. This key observation motivates us to propose Prune&Comp, a novel, plug-and-play iterative layer pruning scheme that leverages magnitude compensation to mitigate such gaps in a training-free manner. Specifically, we first estimate the magnitude gap of layer removal and then eliminate it by rescaling the remaining weights offline. We further demonstrate the advantages of Prune&Comp in improving the stability of iterative pruning. When integrated with an iterative prune-and-compensate loop, Prune&Comp consistently enhances existing layer pruning metrics. For instance, when 5 layers of LLaMA-3-8B are pruned with the prevalent Taylor+ metric, Prune&Comp reduces PPL from 512.78 to 16.34 and retains 90.57% of the original performance across 9 question-answering tasks, outperforming the baseline by 24.72%.

Multivariate Time Series Forecasting (MTSF) aims to capture the dependencies among multiple variables and their temporal dynamics to predict future values. In recent years, Large Language Models (LLMs) have set a new paradigm for MTSF, incorporating external knowledge into the modeling process through textual prompts. However, we observe that current LLM-based methods fail to exploit these priors due to their coarse-grained representation of time series data, which hinders effective alignment of the two modals. To address this, we propose M3Time, a multi-modal, multi-scale, and multi-frequency framework for multivariate time series forecasting. It enhances the quality of time series representations and facilitates the integration of LLM semantic priors with fine-grained temporal features. Additionally, M3Time further improved training stability and model robustness with an adaptive mixed loss function, which dynamically balances L1 and L2 error terms. Experiment results on seven real-world public datasets show that M3Time consistently outperforms state-of-the-art methods, underscoring its effectiveness.

Xiangfeng Wang, Hangyu Guo, Yanlin Lai, Mitt Huang, Liang Zhao, Chengyuan Yao, Yinmin Zhang, Qi Han, Xiaoxiaoren, Chun Yuan, Tong Xu, Zheng Ge, Xiangyu Zhang, Daxin Jiang. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

Data-Free Meta-Learning (DFML) aims to enable efficient learning of unseen few-shot tasks, by meta-learning from multiple pre-trained models without accessing their original training data. While existing DFML methods typically generate synthetic data from these models to perform meta-learning, a comprehensive analysis of DFML's robustness-particularly its failure modes and vulnerability to potential attacks-remains notably absent. Such an analysis is crucial as algorithms often operate in complex and uncertain real-world environments. This paper fills this significant gap by systematically investigating the robustness of DFML, identifying two critical but previously overlooked vulnerabilities: Task-Distribution Shift (TDS) and Task-Distribution Corruption (TDC). TDS refers to the sequential shifts in the evolving task distribution, leading to the catastrophic forgetting of previously learned meta-knowledge. TDC exposes a security flaw of DFML, revealing its susceptibility to attacks when the pre-trained model pool includes untrustworthy models that deceptively claim to be beneficial but are actually harmful. To mitigate these vulnerabilities, we propose a trustworthy DFML framework comprising three components: synthetic task reconstruction, meta-learning with task memory interpolation, and automatic model selection. Specifically, utilizing model inversion techniques, we reconstruct synthetic tasks from multiple pre-trained models to perform meta-learning. To prevent forgetting, we introduce a strategy to replay interpolated historical tasks to efficiently recall previous meta-knowledge. Furthermore, our framework seamlessly incorporates an automatic model selection mechanism to automatically filter out untrustworthy models during the meta-learning process. Extensive experiments across various datasets with two types of untrustworthy models confirm the superiority of our method in significantly enhancing the robustness of DFML.

Recent studies show that the visual place recognition (VPR) method using pre-trained visual foundation models can achieve promising performance. In our previous work, we propose a novel method to realize seamless adaptation of foundation models to VPR (SelaVPR). This method can produce both global and local features that focus on discriminative landmarks to recognize places for two-stage VPR by a parameter-efficient adaptation approach. Although SelaVPR has achieved competitive results, we argue that the previous adaptation is inefficient in training time and GPU memory usage, and the re-ranking paradigm is also costly in retrieval latency and storage usage. In pursuit of higher efficiency and better performance, we propose an extension of the SelaVPR, called SelaVPR++. Concretely, we first design a parameter-, time-, and memory-efficient adaptation method that uses lightweight multi-scale convolution (MultiConv) adapters to refine intermediate features from the frozen foundation backbone. This adaptation method does not back-propagate gradients through the backbone during training, and the MultiConv adapter facilitates feature interactions along the spatial axes and introduces proper local priors, thus achieving higher efficiency and better performance. Moreover, we propose an innovative re-ranking paradigm for more efficient VPR. Instead of relying on local features for re-ranking, which incurs huge overhead in latency and storage, we employ compact binary features for initial retrieval and robust floating-point (global) features for re-ranking. To obtain such binary features, we propose a similarity-constrained deep hashing method, which can be easily integrated into the VPR pipeline. Finally, we improve our training strategy and unify the training protocol of several common training datasets to merge them for better training of VPR models. Extensive experiments show that SelaVPR++ is highly efficient in training time, GPU memory usage, and retrieval latency (6000× faster than TransVPR), as well as outperforms the state-of-the-art methods by a large margin (ranks 1st on MSLS challenge leaderboard).

Given a reference image, composed image retrieval aims to search for the target image from the image gallery with a modification sentence. The complicated semantic relationships within the composed query to comprehend the preserved visual context and modified semantics pose a challenge for the researchers. Most current approaches concentrate on the global or fragmental composition to combine the multi-modal query features and bridge the distance between the query and the target. However, global-level matching tends to ignore fine-grained components that distinguish the matched instance from hard negative instances. As for the fragment-level matching, the over-sensitivity to the local noise due to the lack of explicit local labels and excessive computational demands may have a detrimental impact on performance. In this paper, we propose Semantic Composition via Optimal Transport (SCOT) for composed image retrieval to address the aforementioned issues. To optimize the computational resources, SCOT first introduces Hybrid-modal Affinity Summarization (HAS) to compact salient relationships in the affinity matrix from the reference image patches and modification tokens. Furthermore, to mitigate the over-confidence in local noise, SCOT adaptively assigns optimal transport weights on the salient aligned fragments to Retain Prototype from Alignments (RAS) and guides discriminant semantics learning to Update Semantics from Enhanced Embeddings (USEE), while preserving the characteristics from global structures. Specifically, RAS module investigates the preserved semantics from the reference image based on the affinity matrix, meanwhile USEE module distills significant modification semantics. The proposed SCOT has demonstrated advanced experimental performances on widely-used datasets FashionIQ, CIRR, and Fashion200K. Ablative studies manifest the effectiveness of optimal transport guidance on preserved visual prototypes and updated semantics.

Image restoration aims to restore high-quality images from degraded inputs caused by factors such as motion blur, defocus blur, and rain, where the primary difference between degraded and high-quality images lies in their high-frequency components. Despite the critical role of high frequencies in restoration, few methods explicitly prioritize computational resources for high frequencies over low frequencies. To address this issue, we propose a High-Frequency Prioritized Sparse Attention Network (HFP-SAN), a novel architecture for image restoration tasks. We explicitly prioritize high-frequency components by designing a symmetric encoder-decoder framework integrated with High-Frequency Selective Sparse Attention (HFSSA) modules while handling low-frequency components with a smaller residual network, thereby proportionally allocating computational resources based on their relative importance. HFSSA incorporates a Frequency-Selective Matching (FSM) algorithm to focus attention on strongly correlated high-frequency regions, mitigating computation on areas with weak correlations and irrelevant areas. Additionally, we introduce a dynamically adjustable high-frequency mask that guides the network to focus on the severely degraded regions, further refining restoration quality. The above designs ensure the final reconstructed image is a high-quality product. Experiments demonstrate that our HFP-SAN achieves state-of-the-art performance across multiple image restoration tasks, both quantitatively and qualitatively.

The task of composed image retrieval aims to match the multi-modal query composed of a reference image and a modification sentence with the target image. Most current approaches narrow the distances between the composed queries and targets by investigating matched correspondences in positive triplets. Nevertheless, they are inclined to exhibit heavy reliance on partial correlations. As the negative correspondences are underestimated, semantic clues that distinguish the target from mismatched candidates are obscured by incomplete associations. Moreover, the correlations between the modification textual features and the visual variations from the reference to candidates are imperative to further strengthen the semantic discriminations. In this paper, we propose DIscriminative Perception from NEgative Correspondences (DIPNEC) to address the aforementioned issues. To encourage awareness of the differences between matched and mismatched correspondences, DIPNEC introduces optimal transport with semantic preservation for reassignments on hard negative triplets. Besides, Difference Quantization Alignments (DQA) and Composed Word-level Alignments (CWA) jointly determine the matching scores between multi-modal queries and candidates. Specifically, DQA concentrates on the correlations of textual features with source-to-target visual differences, and CWA further emphasizes the differentiated semantics. DIPNEC has demonstrated competitive performances on the experimental results and ablation studies on widely-used datasets FashionIQ and CIRR.

Zero-shot Referring Image Segmentation (RIS) identifies the instance mask that best aligns with a specified referring expression without training and fine-tuning, significantly reducing the labor-intensive annotation process. Despite achieving commendable results, previous CLIP-based models have a critical drawback: the models exhibit a notable reduction in their capacity to discern relative spatial relationships of objects. This is because they generate all possible masks on an image and evaluate each masked region for similarity to the given expression, often resulting in decreased sensitivity to direct positional clues in text inputs. Moreover, most methods have weak abilities to manage relationships between primary words and their contexts, causing confusion and reduced accuracy in identifying the correct target region. To address these challenges, we propose IteRPrimE (Iterative Grad-CAM Refinement and Primary word Emphasis), which leverages a saliency heatmap through Grad-CAM from a Vision-Language Pre-trained (VLP) model for image-text matching. An iterative Grad-CAM refinement strategy is introduced to progressively enhance the model's focus on the target region and overcome positional insensitivity, creating a self-correcting effect. Additionally, we design the Primary Word Emphasis module to help the model handle complex semantic relations, enhancing its ability to attend to the intended object. Extensive experiments conducted on the RefCOCO/+/g, and PhraseCut benchmarks demonstrate that IteRPrimE outperforms previous SOTA zero-shot methods, particularly excelling in out-of-domain scenarios.

Pseudo-label learning methods have been widely applied in weakly-supervised temporal action localization. Existing works directly utilize weakly-supervised base model to generate instance-level pseudo-labels for training the fully-supervised detection head. We argue that the noise in pseudo-labels would interfere with the learning of fully-supervised detection head, leading to significant performance leakage. Issues with noisy labels include:(1) inaccurate boundary localization; (2) undetected short action clips; (3) multiple adjacent segments incorrectly detected as one segment. To target these issues, we introduce a two-stage noisy label learning strategy to harness every potential useful signal in noisy labels. First, we propose a frame-level pseudo-label generation model with a context-aware denoising algorithm to refine the boundaries. Second, we introduce an online-revised teacher-student framework with a missing instance compensation module and an ambiguous instance correction module to solve the short-action-missing and many-to-one problems. Besides, we apply a high-quality pseudo-label mining loss in our online-revised teacher-student framework to add different weights to the noisy labels to train more effectively. Our model outperforms the previous state-of-the-art method in detection accuracy and inference speed greatly upon the THUMOS14 and ActivityNet v1.2 benchmarks.

Large Language Models (LLMs) such as ChatGPT demonstrate strong few-shot adaptability without requiring fine-tuning, positioning them ideal for data-limited and real-time applications. However, this adaptability has not yet been replicated in current Visual Foundation Models (VFMs), which require explicit fine-tuning with sufficient tuning data. Besides, the pretraining-finetuning paradigm has led to the surge of numerous task-specific modular components, such as Low-Rank Adaptation (LoRA). For the first time, we explore the potential of reusing diverse pre-tuned LoRAs without accessing their original training data, to achieve tuning-free few-shot adaptation in VFMs. Our framework, LoRA Recycle, distills a meta-LoRA from diverse pre-tuned LoRAs with a meta-learning objective, using synthetic data inversely generated from pre-tuned Lo-RAs themselves. The VFM, once equipped with the meta-LoRA, is empowered to solve new few-shot tasks in a single forward pass, akin to the in-context learning of LLMs. Additionally, we incorporate a double-efficient mechanism, accelerating the data-generation and meta-training process while maintaining or even improving performance. Extensive experiments across various few-shot classification benchmarks across both in- and cross-domain scenarios demonstrate the superiority of our framework. Code is available at https://github.com/Egg-Hu/LoRA-Recycle.

The Transformer architecture has revolutionized various fields since it was proposed, where positional encoding plays an essential role in effectively capturing sequential order and context. Therefore, Rotary Positional Encoding (RoPE) was proposed to alleviate these issues, which integrates positional information by rotating the embeddings in the attention mechanism. However, RoPE utilizes manually defined rotation matrices, a design choice that favors computational efficiency but limits the model’s flexibility and adaptability. In this work, we propose ComRoPE, which generalizes RoPE by defining it in terms of trainable commuting angle matrices. Specifically, we demonstrate that pairwise commutativity of these matrices is essential for RoPE to achieve scalability and positional robustness. We formally define the RoPE Equation, which is an essential condition that ensures consistent performance with position offsets. Based on the theoretical analysis, we present two types of trainable commuting angle matrices as sufficient solutions to the RoPE equation, which significantly improve performance, surpassing the current state-of-the-art method by 1.6% at training resolution and 2.9% at higher resolution on the ImageNet-1K dataset. Furthermore, our framework shows versatility in generalizing to existing RoPE formulations and offering new insights for future positional encoding research. To ensure reproducibility, the source code and instructions are available at https://github.com/Longin-Yu/ComRoPE.

Recent breakthroughs in Multimodal Large Language Models (MLLMs) have gained significant recognition within the deep learning community, where the fusion of the Video Foundation Models (VFMs) and Large Language Models(LLMs) has proven instrumental in constructing robust video understanding systems, effectively surmounting constraints associated with predefined visual tasks. These sophisticated MLLMs exhibit remarkable proficiency in comprehending videos, swiftly attaining unprecedented performance levels across diverse benchmarks. However, their operation demands substantial memory and computational resources, underscoring the continued importance of traditional models in video comprehension tasks. In this paper, we introduce a novel learning paradigm termed MLLM4WTAL. This paradigm harnesses the potential of MLLM to offer temporal action key semantics and complete semantic priors for conventional Weakly-supervised Temporal Action Localization (WTAL) methods. MLLM4WTAL facilitates the enhancement of WTAL by leveraging MLLM guidance. It achieves this by integrating two distinct modules: Key Semantic Matching (KSM) and Complete Semantic Reconstruction (CSR). These modules work in tandem to effectively address prevalent issues like incomplete and over-complete outcomes common in WTAL methods. Rigorous experiments are conducted to validate the efficacy of our proposed approach in augmenting the performance of various heterogeneous WTAL models.

Recently, text-guided image editing has achieved significant success. However, existing methods can only apply simple textures like wood or gold when changing the texture of an object. Complex textures such as cloud or fire pose a challenge. This limitation stems from that the target prompt needs to contain both the input image content and, restricting the texture representation. In this paper, we propose TextureDiffusion, a tuning-free image editing method applied to various texture transfer. Initially, the target prompt is directly set to "", making the texture disentangled from the input image content to enhance texture representation. Subsequently, query features in self-attention and features in residual blocks are utilized to preserve the structure of the input image. Finally, to maintain the background, we introduce an edit localization technique which blends the self-attention results and the intermediate latents. Comprehensive experiments demonstrate that TextureDiffusion can harmoniously transfer various textures with excellent structure and background preservation. Code is publicly available at https://github.com/THU-CVML/TextureDiffusion

Recent advancements in multimodal vision models have highlighted limitations in late-stage feature fusion and suboptimal query selection for hybrid prompts open-world segmentation, alongside constraints from caption-derived vocabularies. To address these challenges, we propose Prompt-DINO, a text-guided visual Prompt DINO framework featuring three key innovations. First, we introduce an early fusion mechanism that unifies text/visual prompts and backbone features at the initial encoding stage, enabling deeper cross-modal interactions to resolve semantic ambiguities. Second, we design order-aligned query selection for DETR-based architectures, explicitly optimizing the structural alignment between text and visual queries during decoding to enhance semantic-spatial consistency. Third, we develop a generative data engine powered by the Recognize Anything via Prompting (RAP) model, which synthesizes 0.5B diverse training instances through a dual-path cross-verification pipeline, reducing label noise by 80.5% compared to conventional approaches. Extensive experiments demonstrate that Prompt-DINO achieves state-of-the-art performance on open-world detection benchmarks while significantly expanding semantic coverage beyond fixed-vocabulary constraints. Our work establishes a new paradigm for scalable multimodal detection and data generation in open-world scenarios. Data&Code are available at https://github.com/WeChatCV/WeVisionOne.

Creating high-fidelity 3D meshes with arbitrary topology, including open surfaces and complex interiors, remains a significant challenge. Existing implicit field methods often require costly and detail-degrading watertight conversion, while other approaches struggle with high resolutions. This paper introduces SparseFlex, a novel sparse-structured isosurface representation that enables differentiable mesh reconstruction at resolutions up to $1024^3$ directly from rendering losses. SparseFlex combines the accuracy of Flexicubes with a sparse voxel structure, focusing computation on surface-adjacent regions and efficiently handling open surfaces. Crucially, we introduce a frustum-aware sectional voxel training strategy that activates only relevant voxels during rendering, dramatically reducing memory consumption and enabling high-resolution training. This also allows, for the first time, the reconstruction of mesh interiors using only rendering supervision. Building upon this, we demonstrate a complete shape modeling pipeline by training a variational autoencoder (VAE) and a rectified flow transformer for high-quality 3D shape generation. Our experiments show state-of-the-art reconstruction accuracy, with a ~82% reduction in Chamfer Distance and a ~88% increase in F-score compared to previous methods, and demonstrate the generation of high-resolution, detailed 3D shapes with arbitrary topology. By enabling high-resolution, differentiable mesh reconstruction and generation with rendering losses, SparseFlex significantly advances the state-of-the-art in 3D shape representation and modeling.

Text-to-image generation has greatly advanced content creation, yet accurately rendering visual text remains a key challenge due to blurred glyphs, semantic drift, and limited style control. Existing methods often rely on pre-rendered glyph images as conditions, but these struggle to retain original font styles and color cues, necessitating complex multi-branch designs that increase model overhead and reduce flexibility. To address these issues, we propose a segmentation-guided framework that uses pixel-level visual text masks -- rich in glyph shape, color, and spatial detail -- as unified conditional inputs. Our method introduces two core components: (1) a fine-tuned bilingual segmentation model for precise text mask extraction, and (2) a streamlined diffusion model augmented with adaptive glyph conditioning and a region-specific loss to preserve textual fidelity in both content and style. Our approach achieves state-of-the-art performance on the AnyText benchmark, significantly surpassing prior methods in both Chinese and English settings. To enable more rigorous evaluation, we also introduce two new benchmarks: GlyphMM-benchmark for testing layout and glyph consistency in complex typesetting, and MiniText-benchmark for assessing generation quality in small-scale text regions. Experimental results show that our model outperforms existing methods by a large margin in both scenarios, particularly excelling at small text rendering and complex layout preservation, validating its strong generalization and deployment readiness.

By leveraging the generative priors from pre-trained text-to-image diffusion models, significant progress has been made in real-world image super-resolution (Real-ISR). However, these methods tend to generate inaccurate and unnatural reconstructions in complex and/or heavily degraded scenes, primarily due to their limited perception and understanding capability of the input low-quality image. To address these limitations, we propose, for the first time to our knowledge, to adapt the pre-trained autoregressive multimodal model such as Lumina-mGPT into a robust Real-ISR model, namely PURE, which Perceives and Understands the input low-quality image, then REstores its high-quality counterpart. Specifically, we implement instruction tuning on Lumina-mGPT to perceive the image degradation level and the relationships between previously generated image tokens and the next token, understand the image content by generating image semantic descriptions, and consequently restore the image by generating high-quality image tokens autoregressively with the collected information. In addition, we reveal that the image token entropy reflects the image structure and present a entropy-based Top-k sampling strategy to optimize the local structure of the image during inference. Experimental results demonstrate that PURE preserves image content while generating realistic details, especially in complex scenes with multiple objects, showcasing the potential of autoregressive multimodal generative models for robust Real-ISR. The model and code will be available at https://github.com/nonwhy/PURE.

Multimodal Large Language Models (MLLMs) have emerged as powerful tools for chart comprehension. However, they heavily rely on extracted content via OCR, which leads to numerical hallucinations when chart textual annotations are sparse. While existing methods focus on scaling instructions, they fail to address the fundamental challenge, i.e., reasoning with visual perception. In this paper, we identify a critical observation: MLLMs exhibit weak grounding in chart elements and proportional relationships, as evidenced by their inability to localize key positions to match their reasoning. To bridge this gap, we propose PointCoT, which integrates reflective interaction into chain-of-thought reasoning in charts. By prompting MLLMs to generate bounding boxes and re-render charts based on location annotations, we establish connections between textual reasoning steps and visual grounding regions. We further introduce an automated pipeline to construct ChartPoint-SFT-62k, a dataset featuring 19.2K high-quality chart samples with step-by-step CoT, bounding box, and re-rendered visualizations. Leveraging this data, we develop two instruction-tuned models, ChartPointQ2 and ChartPointQ2.5, which outperform state-of-the-art across several chart benchmarks, e.g., +5.04\% on ChartBench.

Zero-Shot Learning (ZSL) enables the recognition of unseen classes by transferring semantic knowledge from seen classes, typically through shared attributes. Continual Zero-Shot Learning (CZSL) advances the concept by enabling the model to continuously recognize new classes while retaining the ability to recognize previously learned ones. A common approach in CZSL is generative replay, where pseudo-data is synthesized to retain recognition of old classes while adapting to new ones. Despite this, existing methods rely on manually defined semantic prototypes, which may not align well with diverse visual representations. In this paper, we propose a method that refines semantic prototypes in the continual setting, treating them similarly to visual feature generation for better alignment with category semantics. We also introduce a contrastive loss on generated embeddings, which outperforms simplistic loss functions used in prior work. Furthermore, we investigate the impact of SGD (Stochastic Gradient Descent) noise, a factor often overlooked in CZSL, highlighting its significant role in model convergence and generalization. Our contributions improve CZSL performance and enhance the understanding of SGD noise in this context.

Temporal Action Localization (TAL) has garnered significant attention in information retrieval. Existing supervised or weakly supervised methods heavily rely on labeled temporal boundaries and action categories, which are labor-intensive and time-consuming. Consequently, unsupervised temporal action localization (UTAL) has gained popularity. However, current methods face two main challenges: 1) Classification pre-trained features overly focus on highly discriminative regions; 2) Solely relying on visual modality information makes it difficult to determine contextual boundaries. To address these issues, we propose a CLIP-assisted cross-view audio-visual enhanced UTAL method. Specifically, we introduce visual language pre-training (VLP) and classification pre-training-based collaborative enhancement to avoid excessive focus on highly discriminative regions; we also incorporate audio perception to provide richer contextual boundary information. Finally, we introduce a self-supervised cross-view learning paradigm to achieve multi-view perceptual enhancement without additional annotations. Extensive experiments on two public datasets demonstrate our model’s superiority over several state-of-the-art competitors.

Vision-language models such as CLIP have demonstrated strong zero-shot performance, but their considerable size and inefficient inference limit customizable deployment for users. While knowledge distillation is a solution, it still requires the original data, which is not always available due to copyrights and privacy concerns. For many users seeking open-vocabulary customization, Data-Free Knowledge Distillation (DFKD) emerges as a promising direction. Upon rethinking DFKD, we find that existing methods fail on CLIP due to their heavy reliance on BatchNorm layers, which are unexpectedly unusable in CLIP. Based on our findings, we adopt image-text matching to achieve DFKD for CLIP, enabling customization based on arbitrary class texts. This involves (i) inversing a surrogate dataset from CLIP based on text prompts; and (ii) distilling a student model from CLIP using the surrogate dataset. Specifically, we introduce style dictionary diversification to enhance the diversity of synthetic images. To prevent uncontrollable semantics introduced by diversification, we propose a class consistency maintaining strategy to ensure the consistency of synthetic images. Based on synthetic images with various styles, we further propose meta knowledge distillation to train the student model with good generalization ability. Moreover, we introduce a simple yet effective method to enable customization based on few example images. Comprehensive experiments showcase the superiority of our approach across twelve customized tasks, achieving a 9.33\% improvement compared to existing DFKD methods.

Automatic chart understanding is crucial for content comprehension and document parsing. Multimodal Large Language Models (MLLMs) have demonstrated remarkable capabilities in chart understanding through domain-specific alignment and fine-tuning. However, current MLLMs still struggle to provide faithful data and reliable analysis only based on charts. To address it, we propose ChartMoE, which employs the Mixture of Expert (MoE) architecture to replace the traditional linear projector to bridge the modality gap. Specifically, we train several linear connectors through distinct alignment tasks, which are utilized as the foundational initialization parameters for different experts. Additionally, we introduce ChartMoE-Align, a dataset with nearly 1 million chart-table-JSON-code quadruples to conduct three alignment tasks (chart-table/JSON/code). Combined with the vanilla connector, we initialize different experts diversely and adopt high-quality knowledge learning to further refine the MoE connector and LLM parameters. Extensive experiments demonstrate the effectiveness of the MoE connector and our initialization strategy, e.g., ChartMoE improves the accuracy of the previous state-of-the-art from 80.48% to 84.64% on the ChartQA benchmark.

Using extensive training data from SA-1B, the Segment Anything Model (SAM) has demonstrated exceptional generalization and zero-shot capabilities, attracting widespread attention in areas such as medical image segmentation and remote sensing image segmentation. However, its performance in the field of image manipulation detection remains largely unexplored and unconfirmed. There are two main challenges in applying SAM to image manipulation detection: a) reliance on manual prompts, and b) the difficulty of single-view information in supporting cross-dataset generalization. To address these challenges, we develops a cross-view prompt learning paradigm called IMDPrompter based on SAM. Benefiting from the design of automated prompts, IMDPrompter no longer relies on manual guidance, enabling automated detection and localization. Additionally, we propose components such as Cross-view Feature Perception, Optimal Prompt Selection, and Cross-View Prompt Consistency, which facilitate cross-view perceptual learning and guide SAM to generate accurate masks. Extensive experimental results from five datasets (CASIA, Columbia, Coverage, IMD2020, and NIST16) validate the effectiveness of our proposed method.

Visible-Infrared Person Re-Identification (VI-ReID) seeks to match individuals across different modalities. Recent methods focus on discriminative feature extraction and hard sample learning. However, they often suffer from semantic misalignment due to horizontal partitioning in local feature extraction and overlook global hard samples in training. Moreover, the widely used PK Sampler cannot ensure viewpoint balance and diversity. To overcome these limitations, we propose the Semantic Alignment and Hard Sample Retraining (SAHSR) framework. This framework incorporates a Recurrent Semantic Aggregation (RSA) module that progressively aggregates and aligns regional semantics with the help of Modality Alignment loss. Besides, we propose a Confidence-based Hard Sample Retraining (CHSR) strategy that identifies and retrains hard samples to improve the model’s robustness. Additionally, we introduce the Viewpoint-Balanced (VB) Sampler to guarantee a balanced distribution of viewpoints. Extensive experiments on VI-ReID benchmarks demonstrate the significant performance gains of our approach, showing state-of-the-art performance. Code will be available.

Weakly supervised temporal action localization aims to learn to locate actions in videos from video-level or point-level labels, avoiding the need for costly frame-level annotations. Unlike previous work that relies solely on visual modality information, we propose incorporating audio information into the weakly supervised temporal action localization task. While audio-visual localization tasks combine audio and visual information for video localization, temporal action localization often deals with action categories that have weak audio cues. To address this, we propose EAV-Mamba, the first audio-visual perception modeling method based on Mamba. Leveraging Mamba’s powerful audio-visual perception capabilities, we developed modules such as Audio-Perceptive Flow Enhancement, Audio-Perceptive RGB Enhancement, and Audio Self-Perceptive Enhancement. Extensive experiments on two publicly available temporal action localization datasets demonstrate that EAV-Mamba achieves efficient audio-visual perception modeling and state-of-the-art performance in weakly supervised temporal action localization tasks.

Model merging seeks to integrate task-specific expert models into a unified architecture while preserving multi-task generalization capabilities, yet parameter interference between constituent models frequently induces performance degradation. Although prior work has explored many merging strategies, resolving interference without additional data for retraining or test-time computation remains challenging. In this paper, we theoretically demonstrate that the task vectors of the linear layer constitute an approximate linear subspace for its corresponding input. Therefore, we can minimize interference under the guidance of task vectors. Based on this insight, we propose WUDI-Merging (Whoever started the interference shoUld enD It), a simple yet effective model merging method that eliminates interference without any additional data or rescaling coefficients. Comprehensive empirical evaluations across vision and language benchmarks demonstrate our method’s superiority, achieving state-of-the-art performance in data-free model merging scenarios (average 10.9% improvement versus baseline methods) while even outperforming mainstream test-time adaptation approaches by 3.3%, and only very few computing resources are required. The source code and implementation details are available at https://github.com/nathanielyvo/WUDI-Merging.

Knowledge distillation typically minimizes the Kullback–Leibler (KL) divergence between teacher and student logits. However, optimizing the KL divergence can be challenging for the student and often leads to sub-optimal solutions. We further show that gradients induced by KL divergence scale with the magnitude of the teacher logits, thereby diminishing updates on low-probability channels. This imbalance weakens the transfer of inter-class information and in turn limits the performance improvements achievable by the student. To mitigate this issue, we propose a plug-and-play auxiliary ranking loss based on Kendall’s $\tau$ coefficient that can be seamlessly integrated into any logit-based distillation framework. It supplies inter-class relational information while rebalancing gradients toward low-probability channels. We demonstrate that the proposed ranking loss is largely invariant to channel scaling and optimizes an objective aligned with that of KL divergence, making it a natural complement rather than a replacement. Extensive experiments on CIFAR-100, ImageNet, and COCO datasets, as well as various CNN and ViT teacher-student architecture combinations, demonstrate that our plug-and-play ranking loss consistently boosts the performance of multiple distillation baselines.

Reinforcement Learning (RL) has emerged as a powerful tool for neural combinatorial optimization, enabling models to learn heuristics that solve complex problems without requiring expert knowledge. Despite significant progress, existing RL approaches face challenges such as diminishing reward signals and inefficient exploration in vast combinatorial action spaces, leading to inefficiency. In this paper, we propose **Preference Optimization**, a novel method that transforms quantitative reward signals into qualitative preference signals via statistical comparison modeling, emphasizing the superiority among sampled solutions. Methodologically, by reparameterizing the reward function in terms of policy and utilizing preference models, we formulate an entropy-regularized RL objective that aligns the policy directly with preferences while avoiding intractable computations. Furthermore, we integrate local search techniques into the fine-tuning rather than post-process to generate high-quality preference pairs, helping the policy escape local optima. Empirical results on various benchmarks, such as the Traveling Salesman Problem (TSP), the Capacitated Vehicle Routing Problem (CVRP) and the Flexible Flow Shop Problem (FFSP), demonstrate that our method significantly outperforms existing RL algorithms, achieving superior convergence efficiency and solution quality.

Recently, quantization has been widely used for the compression and acceleration of large language models (LLMs). Due to the outliers in LLMs, it is crucial to flatten weights and activations to minimize quantization error with equally spaced quantization points. Prior research explores various pre-quantization transformations to suppress outliers, such as per-channel scaling and Hadamard transformation. However, we observe that these transformed weights and activations can still exhibit steep and dispersed distributions. In this paper, we propose FlatQuant (Fast and Learnable Affine Transformation), a new post-training quantization approach that enhances the flatness of weights and activations. Our approach identifies optimal affine transformations for each linear layer, calibrated in hours via a lightweight objective. To reduce runtime overhead of affine transformation, we apply Kronecker product with two lightweight matrices, and fuse all operations in FlatQuant into a single kernel. Extensive experiments demonstrate that FlatQuant establishes a new state-of-the-art benchmark for quantization. For example, it achieves less than 1\% accuracy drop for W4A4 quantization on the LLaMA-3-70B model, surpassing SpinQuant by 7.5\%. Additionally, it provides up to 2.3x prefill speedup and 1.7x decoding speedup compared to the FP16 model. Code is available at: https://github.com/ruikangliu/FlatQuant.

Merging multiple expert models offers a promising approach for performing multi-task learning without accessing their original data. Existing methods attempt to alleviate task conflicts by sparsifying task vectors or promoting orthogonality among them. However, they overlook the fundamental target of model merging: the merged model performs as closely as possible to task-specific models on respective tasks. We find these methods inevitably discard task-specific information that, while causing conflicts, is crucial for performance. Based on our findings, we frame model merging as a constrained optimization problem ($\textit{i.e.}$, minimizing the gap between the merged model and individual models, subject to the constraint of retaining shared knowledge) and solve it via adaptive projective gradient descent. Specifically, we align the merged model with individual models by decomposing and reconstituting the loss function, alleviating conflicts through $\textit{data-free}$ optimization of task vectors. To retain shared knowledge, we optimize this objective by projecting gradients within a $\textit{shared subspace}$ spanning all tasks. Moreover, we view merging coefficients as adaptive learning rates and propose a task-aware, training-free strategy. Experiments show that our plug-and-play approach consistently outperforms previous methods, achieving state-of-the-art results across diverse architectures and tasks in both vision and NLP domains.

Layer pruning has emerged as a widely used technique for compressing large language models (LLMs). However, existing layer pruning approaches often incur substantial performance degradation. We identify the majority of this degradation to a single yet previously overlooked issue: \textit{the mismatch of activation magnitudes at the pruning interface}. The pre-interface activations exhibit significantly different scales from the post-interface ones, causing the distributional shift as it propagates through the remaining layers. To address this issue, we introduce \textsc{LinearPatch}, a lightweight and plug-and-play technique that fuses two operations into one matrix multiply at the pruning interface: (i) a Hadamard transformation that suppresses massive outliers at particular tokens and (ii) a channel-wise scaling that aligns activation statistics. On LLaMA-3-8B, \textsc{LinearPatch} preserves up to \textbf{94.15\%} of the original model's performance when pruning 5 out of 32 layers, outperforming the previous state of the art by \textbf{4\%}. The patch can be further refined with 5K unlabeled samples via memory-efficient offline distillation, pushing the retention to 95.16\% within only 30 minutes on a single GPU. Code is available at \url{https://github.com/chenxinrui-tsinghua/LinearPatch}.

Chain-of-Thought (CoT) prompting improves the reasoning performance of large language models (LLMs) by encouraging step-by-step thinking. However, CoT-based methods depend on intermediate reasoning steps, which limits scalability and generalization. Recent work explores recursive reasoning, where LLMs reuse internal layers across iterations to refine latent representations without explicit CoT supervision. While promising, these approaches often require costly pretraining and lack a principled framework for how reasoning should evolve across iterations. We address this gap by introducing **Flow Chain-of-Thought (Flow CoT)**, a reasoning paradigm that models recursive inference as a progressive trajectory of latent cognitive states. Flow CoT frames each iteration as a distinct cognitive stage—deepening reasoning across iterations without relying on manual supervision. To realize this, we propose **SCOUT** (*Stepwise Cognitive Optimization Using Teachers*), a lightweight fine-tuning framework that enables Flow CoT-style reasoning without the need for pretraining. SCOUT uses progressive distillation to align each iteration with a teacher of appropriate capacity, and a cross-attention-based retrospective module that integrates outputs from previous iterations while preserving the model’s original computation flow. Experiments across eight reasoning benchmarks show that SCOUT consistently improves both accuracy and explanation quality, achieving up to 1.8\% gains under fine-tuning. Qualitative analyses further reveal that SCOUT enables progressively deeper reasoning across iterations—refining both belief formation and explanation granularity. These results not only validate the effectiveness of SCOUT, but also demonstrate the practical viability of Flow CoT as a scalable framework for enhancing reasoning in LLMs.

Visual place recognition (VPR) is typically regarded as a specific image retrieval task, whose core lies in representing images as global descriptors. Over the past decade, dominant VPR methods (e.g., NetVLAD) have followed a paradigm that first extracts the patch features/tokens of the input image using a backbone, and then aggregates these patch features into a global descriptor via an aggregator. This backbone-plus-aggregator paradigm has achieved overwhelming dominance in the CNN era and remains widely used in transformer-based models. In this paper, however, we argue that a dedicated aggregator is not necessary in the transformer era, that is, we can obtain robust global descriptors only with the backbone. Specifically, we introduce some learnable aggregation tokens, which are prepended to the patch tokens before a particular transformer block. All these tokens will be jointly processed and interact globally via the intrinsic self-attention mechanism, implicitly aggregating useful information within the patch tokens to the aggregation tokens. Finally, we only take these aggregation tokens from the last output tokens and concatenate them as the global representation. Although implicit aggregation can provide robust global descriptors in an extremely simple manner, where and how to insert additional tokens, as well as the initialization of tokens, remains an open issue worthy of further exploration. To this end, we also propose the optimal token insertion strategy and token initialization method derived from empirical studies. Experimental results show that our method outperforms state-of-the-art methods on several VPR datasets with higher efficiency and ranks 1st on the MSLS challenge leaderboard. The code is available at https://github.com/lu-feng/image.

Group Relative Policy Optimization (GRPO) fine-tuning has demonstrated significant enhancements in reasoning tasks. However, it often relies on high quality labeled dataset, which is typically difficult to obtain. To address this challenge, we introduce \textbf{N}oise-\textbf{A}ware \textbf{D}ual-\textbf{R}eward \textbf{O}ptimization (\textbf{NaDRO}) to effectively enhances the training of Large Language Models (LLMs) under noisy or ambiguous supervision. NaDRO operates through two key components: \textbf{(1) Preference-based Outcome Reward (POR)},which makes a principled bias-variance tradeoff, reducing training variance by learning from robust preference rankings instead of overfitting to single-best estimates; and \textbf{(2) Context Perception Reward (CPR) mechanism}, which ensures that LLMs conduct necessary qualitative assessment of the current problem state to foster deeper situational understanding prior to decision-making. To validate our approach in a realistic decision-making testbed, we model classic combinatorial optimization problems like the Traveling Salesman Problem (TSP) and Capacitated Vehicle Routing Problem (CVRP) as Markov Decision Processes, generating training data via cost-limited exploration. Our results demonstrate that the fine-tuned Qwen 7B and Llama 3.1-8B models achieve statistically robust performance, significantly outperforming leading LLM baselines and standard fine-tuning methods on these complex benchmarks. Code is released at \url{https://github.com/microsoft/HeurAgenix/tree/NaDRO}.

The explosive growth of generative video models has amplified the demand for reliable copyright preservation of AI-generated content. Despite its popularity in image synthesis, invisible generative watermarking remains largely underexplored in video generation. To address this gap, we propose Safe-Sora, the first framework to embed graphical watermarks directly into the video generation process. Motivated by the observation that watermarking performance is closely tied to the visual similarity between the watermark and cover content, we introduce a hierarchical coarse-to-fine adaptive matching mechanism. Specifically, the watermark image is divided into patches, each assigned to the most visually similar video frame, and further localized to the optimal spatial region for seamless embedding. To enable spatiotemporal fusion of watermark patches across video frames, we develop a 3D wavelet transform-enhanced Mamba architecture with a novel scanning strategy, effectively modeling long-range dependencies during watermark embedding and retrieval. To the best of our knowledge, this is the first attempt to apply state space models to watermarking, opening new avenues for efficient and robust watermark protection. Extensive experiments demonstrate that Safe-Sora achieves state-of-the- art performance in terms of video quality, watermark fidelity, and robustness, which is largely attributed to our proposals. Code and additional supporting materials are provided in the supplementary.

The interactive task of composed image retrieval aims to retrieve the most relevant images with the bi-modal query, consisting of a reference image and a modification sentence. Despite significant efforts to bridge the heterogeneous gap within the bi-modal query and leverage contrastive learning to reduce the disparity between positive and negative triplets, prior methods often fail to ensure reliable matching due to aleatoric and epistemic uncertainty. Specifically, the aleatoric uncertainty stems from underlying semantic correlations within candidate instances and annotation noise, and the epistemic uncertainty is usually caused by overconfidence in dominant semantic categories. In this paper, we propose Robust UNcertainty Calibration (RUNC) to quantify the uncertainty and calibrate the imbalanced semantic distribution. To mitigate semantic ambiguity in similarity distribution between fusion queries and targets, RUNC maximizes the matching evidence by utilizing a high-order conjugate prior distribution to fit the semantic covariances in candidate samples. With the estimated uncertainty coefficient of each candidate, the target distribution is calibrated to encourage balanced semantic alignment. Additionally, we minimize the ambiguity in the fusion evidence when forming the unified query by incorporating orthogonal constraints on explicit textual embeddings and implicit queries, to reduce the representation redundancy. Extensive experiments and ablation analysis on benchmark datasets FashionIQ and CIRR verify the robustness of RUNC in predicting reliable retrieval results from a large image gallery.

Unsupervised domain adaptation (UDA) for visual question answering (VQA) has attracted research interest. However, with Multi-modal Large Language Models (MLLMs) showing great performance on VQA datasets, UDA for VQA based on MLLMs remains unexplored. To fill this gap, we propose the first systematic approach to Unsupervised Domain Adaptation VQA based on MLLMs (UDAM). First, we introduce semantic context feature alignment and domain query feature alignment, which utilize a single token embedding for each modality to capture contextual domain information from unimodal inputs and conduct coarsegrained feature alignment on it, thus alleviating domain shifts in the unimodal feature space. Second, we propose the novel semantics-guided query feature alignment, which differentiates important domain-specific queries from learnable query outputs and conducts fine-grained feature alignment controlled by a semantics-guided weight map to reduce domain shifts in the cross-modal feature space. Third, we devise a pair-wise domain-aware prompt strategy, which aids UDA by prompting MLLMs to discern the commonality of tasks and the distinctiveness of domains in multi-modal inputs. Extensive experiments demonstrate UDAM's effectiveness in adapting MLLMs to unlabeled new domains.

Neural Architecture Search (NAS) has been extensively studied due to its ability in automatic architecture engineering. Existing NAS methods rely heavily on the gradients and data labels, which either incur immense computational costs or suffer from discretization discrepancy due to the supernet structure. Moreover, the majority of them are limited in generating diverse architectures. To alleviate these issues, in this paper, we propose a novel zero-cost proxy called $\mathsf {MeCo}$MeCo based on the Pearson correlation matrix of the feature maps. Unlike the previous work, the computation of $\mathsf {MeCo}$MeCo as well as its variant $\mathsf {MeCo_{opt}}$MeCoopt requires only one random data for a single forward pass. Based on the proposed zero-cost proxy, we further craft a new zero-shot NAS scheme called $\mathsf {FLASH}$FLASH, which harnesses a new proxy-based operation scoring function and a greedy heuristic. Compared to the existing methods, $\mathsf {FLASH}$FLASH is highly efficient and can construct diverse model architectures instead of repeated cells. We design comprehensive experiments and extensively evaluate our designs on multiple benchmarks and datasets. The experimental results show that our method is one to six orders of magnitudes more efficient than the state-of-the-art baselines with the highest model accuracy.

Visual place recognition (VPR) aims to determine the general geographical location of a query image by retrieving visually similar images from a large geo-tagged database. To obtain a global representation for each place image, most approaches typically focus on the aggregation of deep features extracted from a backbone through using current prominent architectures (e.g., CNNs, MLPs, pooling layer, and transformer encoder), giving little attention to the transformer decoder. However, we argue that its strong capability to capture contextual dependencies and generate accurate features holds considerable potential for the VPR task. To this end, we propose an Efficient Decoder Transformer (EDTformer) for feature aggregation, which consists of several stacked simplified decoder blocks followed by two linear layers to directly produce robust and discriminative global representations. Specifically, we do this by formulating deep features as the keys and values, as well as a set of learnable parameters as the queries. Our EDTformer can fully utilize the contextual information within deep features, then gradually decode and aggregate the effective features into the learnable queries to output the global representations. Moreover, to provide more powerful deep features for EDTformer and further facilitate the robustness, we use the foundation model DINOv2 as the backbone and propose a Low-rank Parallel Adaptation (LoPA) method to enhance its performance in VPR, which can refine the intermediate features of the backbone progressively in a memory- and parameter-efficient way. As a result, our method not only outperforms single-stage VPR methods on multiple benchmark datasets, but also outperforms two-stage VPR methods which add a re-ranking with considerable cost. Code will be available at https://github.com/Tong-Jin01/EDTformer.

Long-Tailed Recognition (LTR) poses significant challenges due to the heavily imbalanced nature of real-world data, which severely skews data-driven deep neural networks. Despite the rapid progress of Vision-Language Models (VLMs), they still face challenges in effectively learning from long-tailed visual data. In this paper, we present a comprehensive analysis of the reasons behind the underperformance of VLMs and propose a hierarchical inference framework to address this issue. Specifically, we prompt the large language models to generatesentence-leveldescriptors for class labels and conduct the open vocabulary classification by computing the average similarity between the image and each descriptor. Areweightingmechanism is further proposed to filter out uninformative descriptors. To mitigate model bias incurred by the long-tail distribution, we propose a feature adapter with the logit adjustment technique and fine-tune the CLIP model via visual prompt tokens. We introduce the Shared Feature space Mixup (SFM) to enhance the interaction between modalities to address tail visual feature insufficiency. Finally, we propose a hierarchical inference manner to combine the aforementioned proposals. Extensive evaluations demonstrate that our approach achieves state-of-the-art performance by fine-tuning only a few parameters on the Places-LT, ImageNet-LT, and iNaturalist 2018 benchmarks.

Many current image restoration approaches utilize neural networks to acquire robust image-level priors from extensive datasets, aiming to reconstruct missing details. Nevertheless, these methods often falter with images that exhibit significant information gaps. While incorporating external priors or leveraging reference images can provide supplemental information, these strategies are limited in their practical scope. Alternatively, textual inputs offer greater accessibility and adaptability. In this study, we develop a sophisticated framework enabling users to guide the restoration of deteriorated images via textual descriptions. Utilizing the text-image compatibility feature of CLIP enhances the integration of textual and visual data. Our versatile framework supports multiple restoration activities such as image inpainting, super-resolution, and colorization. Comprehensive testing validates our technique's efficacy.

Blind face restoration under extreme conditions involves reconstructing high-quality face images from severely degraded inputs. These input images are often in poor quality and have extreme facial poses, leading to errors in facial structure and unnatural artifacts within the restored images. In this paper, we show that utilizing 3D priors effectively compensates for structure knowledge deficiencies in 2D priors while preserving the texture details. Based on this, we introduce FREx (Face Restoration under Extreme conditions) that combines structure-accurate 3D priors and texture-rich 2D priors in pretrained generative networks for blind face restoration under extreme conditions. To fuse the different information in 3D and 2D priors, we introduce an adaptive weight module that adjusts the importance of features based on the input image's condition. With this approach, our model can restore structure-accurate and natural-looking faces even when the images have lost a lot of information due to degradation and extreme pose. Extensive experimental results on synthetic and real-world datasets validate the effectiveness of our methods.

Visual place recognition (VPR) is a fundamental task for many applications such as robot localization and augmented reality. Recently, the hierarchical VPR methods have received considerable attention due to the trade-off between accuracy and efficiency. They usually first use global features to retrieve the candidate images, then verify the spatial consistency of matched local features for re-ranking. However, the latter typically relies on the RANSAC algorithm for fitting homography, which is time-consuming and non-differentiable. This makes existing methods compromise to train the network only in global feature extraction. Here, we propose a transformer-based deep homography estimation (DHE) network that takes the dense feature map extracted by a backbone network as input and fits homography for fast and learnable geometric verification. Moreover, we design a re-projection error of inliers loss to train the DHE network without additional homography labels, which can also be jointly trained with the backbone network to help it extract the features that are more suitable for local matching. Extensive experiments on benchmark datasets show that our method can outperform several state-of-the-art methods. And it is more than one order of magnitude faster than the mainstream hierarchical VPR methods using RANSAC. The code is released at https://github.com/Lu-Feng/DHE-VPR.

Unsupervised domain adaptation object detection(UDAOD) research on Detection Transformer(DETR) mainly focuses on feature alignment and existing methods can be divided into two kinds, each of which has its unresolved issues. One-stage feature alignment methods can easily lead to performance fluctuation and training stagnation. Two-stage feature alignment method based on mean teacher comprises a pretraining stage followed by a self-training stage, each facing problems in obtaining reliable pretrained model and achieving consistent performance gains. Methods mentioned above have not yet explore how to utilize the third related domain such as target-like domain to assist adaptation. To address these issues, we propose a two-stage framework named MTM, i.e. Mean Teacher-DETR with Masked Feature Alignment. In the pretraining stage, we utilize labeled target-like images produced by image style transfer to avoid performance fluctuation. In the self-training stage, we leverage unlabeled target images by pseudo labels based on mean teacher and propose a module called Object Queries Knowledge Transfer(OQKT) to ensure consistent performance gains of the student model. Most importantly, we propose masked feature alignment methods including Masked Domain Query-based Feature Alignment(MDQFA) and Masked Token-wise Feature Alignment(MTWFA) to alleviate domain shift in a more robust way, which not only prevent training stagnation and lead to a robust pretrained model in the pretraining stage, but also enhance the model's target performance in the self-training stage. Experiments on three challenging scenarios and a theoretical analysis verify the effectiveness of MTM.

Image super-resolution is a fundamentally ill-posed problem because multiple valid high-resolution images exist for one low-resolution image. Super-resolution methods based on diffusion probabilistic models can deal with the ill-posed nature by learning the distribution of high-resolution images conditioned on low-resolution images, avoiding the problem of blurry images in PSNR-oriented methods. However, existing diffusion-based super-resolution methods have high time consumption with the use of iterative sampling, while the quality and consistency of generated images are less than ideal due to problems like color shifting. In this paper, we propose Efficient Conditional Diffusion Model with Probability Flow Sampling (ECDP) for image super-resolution. To reduce the time consumption, we design a continuous-time conditional diffusion model for image super-resolution, which enables the use of probability flow sampling for efficient generation. Additionally, to improve the consistency of generated images, we propose a hybrid parametrization for the denoiser network, which interpolates between the data-predicting parametrization and the noise-predicting parametrization for different noise scales. Moreover, we design an image quality loss as a complement to the score matching loss of diffusion models, further improving the consistency and quality of super-resolution. Extensive experiments on DIV2K, ImageNet, and CelebA demonstrate that our method achieves higher super-resolution quality than existing diffusion-based image super-resolution methods while having lower time consumption. Our code is available at https://github.com/Yuan-Yutao/ECDP.

Over the past decade, most methods in visual place recognition (VPR) have used neural networks to produce feature representations. These networks typically produce a global representation of a place image using only this image itself and neglect the cross-image variations (e.g. viewpoint and illumination), which limits their robustness in challenging scenes. In this paper, we propose a robust global representation method with cross-image correlation awareness for VPR, named CricaVPR. Our method uses the attention mechanism to correlate multiple images within a batch. These images can be taken in the same place with different conditions or viewpoints, or even captured from different places. Therefore, our method can utilize the cross-image variations as a cue to guide the representation learning, which ensures more robust features are produced. To further facilitate the robustness, we propose a multi-scale convolution-enhanced adaptation method to adapt pre-trained visual foundation models to the VPR task, which introduces the multi-scale local information to further enhance the cross-image correlation-aware representation. Experimental results show that our method out-performs state-of-the-art methods by a large margin with significantly less training time. The code is released at https://github.com/Lu-Feng/CricaVPR.

Data-Free Meta-Learning (DFML) aims to extract knowledge from a collection of pre-trained models without requiring the original data, presenting practical benefits in contexts constrained by data privacy concerns. Current DFML methods primarily focus on the data recovery from these pre-trained models. However, they suffer from slow recovery speed and overlook gaps inherent in heterogeneous pre-trained models. In response to these challenges, we introduce the Faster and Better Data-Free Meta-Learning (FREE) framework, which contains: (i) a meta-generator for rapidly recovering training tasks from pre-trained models; and (ii) a meta-learner for generalizing to new unseen tasks. Specifically, within the module Faster Inversion via Meta-Generator, each pre-trained model is perceived as a distinct task. The meta-generator can rapidly adapt to a specific task in just five steps, significantly accelerating the data recovery. Furthermore, we propose Better Generalization via Meta-Learner and introduce an implicit gradient alignment algorithm to optimize the meta-learner. This is achieved as aligned gradient directions alleviate potential conflicts among tasks from heterogeneous pre-trained models. Empirical experiments on multiple benchmarks affirm the superiority of our approach, marking a notable speed-up (20x) and performance enhancement (1.42% ~ 4.78%) in comparison to the state-of-the-art.

In image restoration (IR), leveraging semantic priors from segmentation models has been a common approach to improve performance. The recent segment anything model (SAM) has emerged as a powerful tool for extracting advanced semantic priors to enhance IR tasks. However, the computational cost of SAM is prohibitive for IR, compared to existing smaller IR models. The incorporation of SAMfor extracting semantic priors considerably hampers the model inference efficiency. To address this issue, we propose a general framework to distill SAM's semantic knowledge to boost exiting IR models without interfering with their inference process. Specifically, our proposed framework consists of the semantic priors fusion (SPF) scheme and the semantic priors distillation (SPD) scheme. SPF fuses two kinds of information between the restored image predicted by the original IR model and the semantic mask predicted by SAM for the refined restored image. SPD leverages a self-distillation manner to distill the fused semantic priors to boost the performance of original IR models. Additionally, we design a semantic-guided relation (SGR) module for SPD, which ensures semantic feature representation space consistency to fully distill the priors. We demonstrate the effectiveness of our framework across multiple IR models and tasks, including deraining, deblurring, and denoising.

Recent studies show that vision models pre-trained in generic visual learning tasks with large-scale data can provide useful feature representations for a wide range of visual perception problems. However, few attempts have been made to exploit pre-trained foundation models in visual place recognition (VPR). Due to the inherent difference in training objectives and data between the tasks of model pre-training and VPR, how to bridge the gap and fully unleash the capability of pre-trained models for VPR is still a key issue to address. To this end, we propose a novel method to realize seamless adaptation of pre-trained models for VPR. Specifically, to obtain both global and local features that focus on salient landmarks for discriminating places, we design a hybrid adaptation method to achieve both global and local adaptation efficiently, in which only lightweight adapters are tuned without adjusting the pre-trained model. Besides, to guide effective adaptation, we propose a mutual nearest neighbor local feature loss, which ensures proper dense local features are produced for local matching and avoids time-consuming spatial verification in re-ranking. Experimental results show that our method outperforms the state-of-the-art methods with less training data and training time, and uses about only 3% retrieval runtime of the two-stage VPR methods with RANSAC-based spatial verification. It ranks 1st on the MSLS challenge leaderboard (at the time of submission). The code is released at https://github.com/Lu-Feng/SelaVPR.

The Segment-Anything Model (SAM) stands as a foundational framework for image segmentation. While it exhibits remarkable zero-shot generalization in typical scenarios, its advantage diminishes when applied to specialized domains like medical imagery and remote sensing. To address this limitation, this paper introduces Conv-LoRA, a simple yet effective parameter-efficient fine-tuning approach. By integrating ultra-lightweight convolutional parameters into Low-Rank Adaptation (LoRA), Conv-LoRA can inject image-related inductive biases into the plain ViT encoder, further reinforcing SAM’s local prior assumption. Notably, Conv-LoRA not only preserves SAM’s extensive segmentation knowledge but also revives its capacity of learning high-level image semantics, which is constrained by SAM’s foreground-background segmentation pretraining. Comprehensive experimentation across diverse benchmarks spanning multiple domains underscores Conv-LoRA’s superiority in adapting SAM to real-world semantic segmentation tasks.

Model inversion, which aims to reconstruct the original training data from pre-trained discriminative models, is especially useful when the original training data is unavailable due to privacy, usage rights, or size constraints. However, existing dense inversion methods attempt to reconstruct the entire image area, making them extremely inefficient when inverting high-resolution images from large-scale Vision Transformers (ViTs). We further identify two underlying causes of this inefficiency: the redundant inversion of noisy backgrounds and the unintended inversion of spurious correlations—a phenomenon we term “hallucination” in model inversion. To address these limitations, we propose a novel sparse model inversion strategy, as a plug-and-play extension to speed up existing dense inversion methods with no need for modifying their original loss functions. Specifically, we selectively invert semantic foregrounds while stopping the inversion of noisy backgrounds and potential spurious correlations. Through both theoretical and empirical studies, we validate the efficacy of our approach in achieving significant inversion acceleration (up to $\times$3.79) while maintaining comparable or even enhanced downstream performance in data-free model quantization and data-free knowledge transfer. Code is available at https://github.com/Egg-Hu/SMI.

Knowledge distillation is a widely adopted model compression technique that has been successfully applied to object detection. In feature distillation, it is common practice for the student model to imitate the feature responses of the teacher model, with the underlying objective of improving its own abilities by reducing the disparity with the teacher. However, it is crucial to recognize that the disparities between the student and teacher are inconsistent, highlighting their varying abilities. In this paper, we explore the inconsistency in the disparity between teacher and student feature maps and analyze their impact on the efficiency of the distillation. We find that regions with varying degrees of difference should be treated separately, with different distillation constraints applied accordingly. We introduce our distillation method called Disparity Feature Distillation(DFD). The core idea behind DFD is to apply different treatments to regions with varying learning difficulties, simultaneously incorporating leniency and strictness. It enables the student to better assimilate the teacher’s knowledge. Through extensive experiments, we demonstrate the effectiveness of our proposed DFD in achieving significant improvements. For instance, when applied to detectors based on ResNet50 such as RetinaNet, FasterRCNN, and RepPoints, our method enhances their mAP from 37.4%, 38.4%, 38.6% to 41.7%, 42.4%, 42.7%, respectively. Our approach also demonstrates substantial improvements on YOLO and ViT-based models. The code is available at https://github.com/luckin99/DFD.

Recent vision-language models have achieved tremendous advances. However, their computational costs are also escalating dramatically, making model acceleration exceedingly critical. To pursue more efficient vision-language Transformers, this paper introduces Cross-Guided Ensemble of Tokens (CrossGET), a general acceleration framework for vision-language Transformers. This framework adaptively combines tokens in real-time during inference, significantly reducing computational costs while maintaining high performance. CrossGET features two primary innovations: 1) Cross-Guided Matching and Ensemble. CrossGET leverages cross-modal guided token matching and ensemble to effectively utilize cross-modal information, achieving wider applicability across both modality-independent models, e.g., CLIP, and modality-dependent ones, e.g., BLIP2. 2) Complete-Graph Soft Matching. CrossGET introduces an algorithm for the token-matching mechanism, ensuring reliable matching results while facilitating parallelizability and high efficiency. Extensive experiments have been conducted on various vision-language tasks, such as image-text retrieval, visual reasoning, image captioning, and visual question answering. The performance on both classic multimodal architectures and emerging multimodal LLMs demonstrates the framework’s effectiveness and versatility. The code is available at https://github.com/sdc17/CrossGET.

Data-Free Meta-Learning (DFML) aims to derive knowledge from a collection of pre-trained models without accessing their original data, enabling the rapid adaptation to new unseen tasks. Current methods often overlook the heterogeneity among pre-trained models, which leads to performance degradation due to task conflicts. In this paper, we empirically and theoretically identify and analyze the model heterogeneity in DFML. We find that model heterogeneity introduces a heterogeneity-homogeneity trade-off, where homogeneous models reduce task conflicts but also increase the overfitting risk. Balancing this trade-off is crucial for learning shared representations across tasks. Based on our findings, we propose Task Groupings Regularization, a novel approach that benefits from model heterogeneity by grouping and aligning conflicting tasks. Specifically, we embed pre-trained models into a task space to compute dissimilarity, and group heterogeneous models together based on this measure. Then, we introduce implicit gradient regularization within each group to mitigate potential conflicts. By encouraging a gradient direction suitable for all tasks, the meta-model captures shared representations that generalize across tasks. Comprehensive experiments showcase the superiority of our approach in multiple benchmarks, effectively tackling the model heterogeneity in challenging multi-domain and multi-architecture scenarios.

Diffusion-based generative model currently serves as a mainstream generative method. The noise schedule has a significant impact on the training process of diffusion model, as it affects both the distribution of the noisy training set and the weights of the objective function at each noise level. In this paper, we design the noise schedule from the scope of reducing the final error upper bound of the reverse denoising process. By examining Monte Carlo training from a theoretical perspective, we establish an association between noise schedule and the upper bound of network output error. Furthermore, we derive the connection between network output and final error through reverse process. We design our noise schedule with the goal of reducing the upper bound of error combined with the correlation analysis of network output. Experimental results demonstrate that our noise schedule enhances perceptual quality on CIFAR-10, FFHQ-64x64 and AFHQv2-64x64. Our noise schedule achieves state-of-the-art FID score of 1.70 on CIFAR-10 unconditional generation task using discriminator guidance method. On FFHQ/AFHQv2, using our noise schedule to retrain the pre-trained model can improve the sample quality at little training cost.

We present a simple yet effective method for skeleton-agnostic motion retargeting. Previous methods transfer motion between high-resolution meshes, failing to preserve the inherent local-part motions in the mesh. Addressing this issue, our proposed method learns the correspondence in a coarse-to-fine fashion by disentangling the retargeting process within multi-scale meshes. First, we propose a mesh-pooling module that pools the mesh representations for better motion transfer. This module improves the ability to handle small-part motion and preserves the local motion interdependence between neighboring mesh vertices. Furthermore, we leverage a multi-scale refinement procedure to complement missing mesh details by gradually refining the low-resolution mesh output with a higher-resolution one. We evaluate our method on several well-known 3D character datasets, and it yields an average improvement of 25% on point-wise mesh Euclidean distance (PMD) against the start-of-art method. Qualitative results show that our method is significantly helpful in preserving the moving consistency of different body parts on the target character due to disentangling body-part structures and mesh details in a multi-scale way.

We propose an effective hybrid approach jointly leveraging local and global features for shortest-path (SP) distance estimation in domain-agnostic large-scale graphs. Previous works struggle to make estimations either from node-wise local embeddings or by compressing a global SP distance matrix, causing insufficient learning at some distance and loss of accuracy. Unlike them, we find a way to better preserve local distance on node embeddings, and then integrate them with a global process for accurate estimation at every distance. First, we propose a distance-consistent embedding method that better preserves the distance between each node and its local neighbors due to resampling node occurrence on random walks. Second, we train a feed-forward network with boosting techniques (FFN-BT) to estimate SP distance from these embeddings plus existing global features. Experimental results show that our approach averagely yields 10% improved accuracy and 20% reduced time when compared to existing methods on a broad class of graphs.

The remarkable performance of large-scale Text-to-Image generation(TI) models is evident in their ability to produce high-quality and diverse images. However, despite advancements, the field of image editing still faces challenges. Current methods struggle to strike a balance between fidelity and powerful editing capabilities. Moreover, approaches that do not involve fine-tuning fail to produce diverse editing results. We introduce Noise Weighting Phased Prompt Image Editing (NWPP), a method that excels in powerful editing, high fidelity, and diverse results without fine-tuning. Our approach involves a two-phase generation process. The first phase employs the original prompt to guide initial image editing, ensuring a layout resembling the original image. In the second phase, a noise-weighting technique based on the Cross-Attention map minimizes the impact of the target text on non-editing regions. Further enhancement is achieved through the integration of the KV injection module, expanding the editing capabilities and enabling diverse result generation. Experimental evaluations, conducted on both generated images and the COCO dataset, affirm the efficacy of our method.

Neural network-based stereo matching algorithmms have made significant progress in fields such as robot navigation and autonomous driving. These application scenarios where fast and accurate obtaining the disparity of stereo images is critical for real-time stereo matching decisions. However, current stereo matching algorithms face the challenge of balancing real-time and accuracy while maintaining high accuracy. In this paper, we propose a disparity update strategy based on geometric encoding (MDStereo), which uses global and non-local geometric feature information to update disparity for highly accurate and quick matching. The proposed MDStereo constructs a group of geometric encoding volumes to encode the local information of the image; Next, a new GEDU method for disparity updating is proposed, which retrieves the correlation of high-resolution cost volumes in the form of sampling, and then fuses the geometric encoding information to iteratively update the disparity. Compared to RAFT-Stereo which retrieves correlations from all cascade cost volumes, our GEDU not only provides rich information but also has a more concise architecture. Furthermore, to speed up the inference of the algorithm, we improve the 3D stacked hourglass network, which effectively increases the receptive field and reduces the computational complexity. Our MDStereo has validated its effectiveness and accuracy on several benchmarks, achieving an EPE (end point error) of 0.58 pixels, a 3-pixel error of 2.58%, and a runtime of 43ms on the Scene Flow dataset. At the time of writing, MDStereo outperformed the published real-time methods at the popular KITTI 2012 and KITTI 2015. Compared with existing iteratively updating disparity methods (e.g., RAFT-Stereo), our method reduces the memory consumption by 54% and greatly improves the inference speed.

The challenging task composed image retrieval targets at identifying the matched image from the multi-modal query with a reference image and a textual modifier. Most existing methods are devoted to composing the unified query representations from the query images and texts, yet the distribution gaps between the hybrid-modal query representations and visual target representations are neglected. However, directly incorporating target features on the query may cause ambiguous rankings and poor robustness due to the insufficient exploration of the distinguishments and overfitting issues. To address the above concerns, we propose a novel framework termed SemAntic Distillation from Neighborhood (SADN) for composed image retrieval. For mitigating the distribution divergences, we construct neighborhood sampling from the target domain for each query and aggregate neighborhood features with adaptive weights to restructure the query representations. Specifically, the adaptive weights are determined by the collaboration of two individual modules, as correspondence-induced adaption and divergence-based correction. Correspondence-induced adaption accounts for capturing the correlation alignments from neighbor features under the guidance of the positive representations, and the divergence-based correction regulates the weights based on the embedding distances between hard negatives and the query in the latent space. Extensive results and ablation studies on CIRR and FashionIQ validate that the proposed semantic distillation from neighborhood significantly outperforms baseline methods.

Incorporating a customized object into image generation presents an attractive feature in text-to-image (T2I) generation. Some methods finetune T2I models for each object individually at test-time, which tend to be overfitted and time-consuming. Others train an extra encoder to extract object visual information for customization efficiently but struggle to preserve the object's identity. To address these limitations, we present CustomNet, a unified encoder-based object customization framework that explicitly incorporates 3D novel view synthesis capabilities into the customization process. This integration facilitates the adjustment of spatial positions and viewpoints, producing diverse outputs while effectively preserving the object's identity. To train our model effectively, we propose a dataset construction pipeline to better handle real-world objects and complex backgrounds. Additionally, we introduce delicate designs that enable location control and flexible background control through textual descriptions or user-defined backgrounds. Our method allows for object customization without the need of test-time optimization, providing simultaneous control over viewpoints, location, and text. Experimental results show that our method outperforms other customization methods regarding identity preservation, diversity, and harmony. Codes are available at https://github.com/TencentARC/CustomNet.

Visual place recognition (VPR) is an essential task for multiple applications such as augmented reality and robot localization. Over the past decade, mainstream methods in the VPR area have been to use feature representation based on global aggregation, as exemplified by NetVLAD. These features are suitable for large-scale VPR and robust against viewpoint changes. However, the VLAD-based aggregation methods usually learn a large number of (e.g., 64) clusters and their corresponding cluster centers, which directly leads to a high dimension of the yielded global features. More importantly, when there is a domain gap between the data in training and inference, the cluster centers determined on the training set are usually improper for inference, resulting in a performance drop. To this end, we first attempt to improve NetVLAD by removing the cluster center and setting only a small number of (e.g., only 4) clusters. The proposed method not only simplifies NetVLAD but also enhances the generalizability across different domains. We name this method SuperVLAD. In addition, by introducing ghost clusters that will not be retained in the final output, we further propose a very low-dimensional 1-Cluster VLAD descriptor, which has the same dimension as the output of GeM pooling but performs notably better. Experimental results suggest that, when paired with a transformer-based backbone, our SuperVLAD shows better domain generalization performance than NetVLAD with significantly fewer parameters. The proposed method also surpasses state-of-the-art methods with lower feature dimensions on several benchmark datasets. The code is available at https://github.com/lu-feng/SuperVLAD.

We delve into the realm of offline meta-reinforcement learning (OMRL), a practical paradigm in the field of reinforcement learning that leverages offline data to adapt to new tasks. While prior approaches have not explored the utilization of context-based dynamical models to tackle OMRL problems, our research endeavors to fill this gap. Our investigation uncovers shortcomings in existing context-based methods, primarily related to distribution shifts during offline learning and challenges in establishing stable task representations. To address these issues, we formulate the problem as Hidden-Parameter MDPs and propose a framework for effective model adaptation using meta-models plus latent variables, which is inferred by the transformer-based system recognition module trained in an unsupervised fashion. Through extensive experimentation encompassing diverse simulated robotics and control tasks, we validate the efficacy of our approach and demonstrate its superior generalization ability compared to existing schemes, and explore multiple strategies for obtaining policies with personalized models. Our method achieves a model with reduced prediction error, outperforming previous methods in policy performance, and facilitating efficient adaptation when compared to prior dynamic model generalization methods and OMRL algorithms.

This article offers an integrated solution for first-person shooter (FPS) games to train agents with adaptive strategies. Solving such complex decision tasks requires generalization ability and adaptive strategies. We develop a framework using a novel adaptive strategic control algorithm combined with advanced techniques, such as the hindsight experience replay, multiagent reinforcement learning, and league training. The approach adopts a multistage learning scheme, consisting of learning a goal-conditioned navigation policy, then transferring to learn sophisticated shooting skills by playing against a league of players, and finally learning adaptive strategies. Our agent achieves the SOTA result in pastViZDoomAI Competitions, surpassing previous top-ranked agents (never seen during training) by a large margin. We provide comprehensive analysis and experiments to elaborate the effect of each component in affecting the agent performance and demonstrate that the proposed and adopted techniques are essential to achieve superior performance inViZDoomCompetition and potentially valuable for general end-to-end FPS games.

Although few-shot learning aims to address data scarcity, it still requires large, annotated datasets for training, which are often unavailable due to cost and privacy concerns. Previous studies have utilized pre-trained diffusion models, either to synthesize auxiliary data besides limited labeled samples, or to employ diffusion models as zero-shot classifiers. However, they are limited to conditional diffusion models needing class prior information (e.g., carefully crafted text prompts) about unseen tasks. To overcome this, we leverage unconditional diffusion models without needs for class information to train a meta-model capable of generalizing to unseen tasks. The framework contains(1)a meta-learning without data approach that uses synthetic data during training; and(2)a diffusion model-based data augmentation to calibrate the distribution shift during testing. During meta-training, we implement aself-taughtclass-learner to gradually capture class concepts, guiding unconditional diffusion models to generate alabeledpseudo dataset. This pseudo dataset is then used to jointly train the class-learner and the meta-model, allowing for iterative refinement and clear differentiation between classes. During meta-testing, we introduce a data augmentation that employs the diffusion models used in meta-training, to narrow the gap between meta-training and meta-testing task distribution. This enables the meta-model trained onsyntheticimages to effectively classifyrealimages in unseen tasks. Comprehensive experiments showcase the superiority and adaptability of our approach in four real-world scenarios. Code available athttps://github.com/WalkerWorldPeace/MLWDUDM.

In unsupervised domain adaptation (UDA), negative transfer is one of the most challenging problems. Due to complex environments, the used domain data are always corrupted by noise or outliers in many applications. If the noisy data are directly used for domain adaptation, the disturbances and negative influence of the noise are also shifted for the target tasks. Thus, preventing disturbances and negative effects caused by noise are key problems in UDA that need to be addressed. In this article, a low-rank correlation learning (LRCL) method is proposed for UDA. In LRCL, the noisy domain data are recovered by low-rank learning; then both domain data are cleaned. Hence, the disturbances and negative effects of the noise are prevented. The maximized correlated features of the clean data from the source and target domains are learned by a novel correlation regularization term in a latent common space. LRCL also reduces the distribution difference of the learned clean source and target data by constructing a reconstruction term, in which the clean target data are linearly represented by the clean source data. To explore the temporal and structural information of the data, we further extend LRCL into a graph case and propose graph LRCL (GLRCL). Extensive experiments have been conducted on several public data benchmarks, and the experimental results demonstrate that our methods can effectively prevent negative transfer and obtain better classification outcomes than other compared approaches.

Composed image retrieval is a challenging task in the field of multi-modal learning, aiming at measuring the similarities between target images and query images with modification sentences. Most previous methods either construct feature composition for the query image and modification text or concentrate on extracting cross-modal alignments. However, these methods are prone to neglect the negative impacts of the mismatched correspondences between the hybrid-modal query and target, which could be discriminative when comparing similar instances. Besides, localized textual representations are not fully explored when learning similarities between the query and the target. To overcome the above issues, we propose a Negative-Sensitive Framework with Semantic Enhancement (NSFSE) for mining the adaptive boundaries between matched and mismatched samples with comprehensive consideration of positive and negative correspondences. It can optimize the threshold dynamically based on distributions to explore the intrinsic characteristics of positive and negative correlations, which could further facilitate accurate similarity learning. A text-guided attention mechanism after infusing cross-modal affinities on localized word features is exploited in NSFSE to explore latent semantic-related visual similarity and cross-modal similarity simultaneously. The performance of extensive experiments and comprehensive analysis on three representative datasets CIRR, FashionIQ, and Fashion200 K demonstrate the effectiveness of negative mining of similarity with semantic enhancement in the proposed NSFSE.

Real-world data usually suffers from severe class imbalance and long-tailed distributions, where minority classes are significantly underrepresented compared to the majority ones. Recent research prefers to utilize multi-expert architectures to mitigate the model uncertainty on the minority, where collaborative learning is employed to aggregate the knowledge of experts, i.e., online distillation. In this article, we observe that the knowledge transfer between experts is imbalanced in terms of class distribution, which results in limited performance improvement of the minority classes. To address it, we propose a re-weighted distillation loss by comparing two classifiers' predictions, which are supervised by online distillation and label annotations, respectively. We also emphasize that feature-level distillation will significantly improve model performance and increase feature robustness. Finally, we propose an Effective Collaborative Learning (ECL) framework that integrates a contrastive proxy task branch to further improve feature quality. Quantitative and qualitative experiments on four standard datasets demonstrate that ECL achieves state-of-the-art performance and the detailed ablation studies manifest the effectiveness of each component in ECL.

With the increase in the number of image data and the lack of corresponding labels, weakly supervised learning has drawn a lot of attention recently in computer vision tasks, especially in the fine-grained semantic segmentation problem. To alleviate human efforts from expensive pixel-by-pixel annotations, our method focuses on weakly supervised semantic segmentation (WSSS) with image-level labels, which are much easier to obtain. As a considerable gap exists between pixel-level segmentation and image-level labels, how to reflect the image-level semantic information on each pixel is an important question. To explore the congeneric semantic regions from the same class to the maximum, we construct the patch-level semantic augmentation network (PatchNet) based on the self-detected patches from different images that contain the same class labels. Patches can frame the objects as much as possible and include as little background as possible. The patch-level semantic augmentation network that is established with patches as the nodes can maximize the mutual learning of similar objects. We regard the embedding vectors of patches as nodes and use a transformer-based complementary learning module to construct weighted edges according to the embedding similarity between different nodes. Moreover, to better supplement semantic information, we propose softcomplementary loss functions matched with the whole network structure. We conduct experiments on the popular PASCAL VOC 2012 and MS COCO 2014 benchmarks, and our model yields the state-of-the-art performance.

Learning with noisy label is a classic problem that has been extensively studied for image tasks, but much less for video in the literature. A straightforward migration from images to videos without considering temporal semantics and computational cost is not a sound choice. In this paper, we propose two new strategies for video analysis with noisy labels: 1) a lightweight channel selection method dubbed as Channel Truncation for feature-based label noise detection. This method selects the most discriminative channels to split clean and noisy instances in each category. 2) A novel contrastive strategy dubbed as Noise Contrastive Learning, which constructs the relationship between clean and noisy instances to regularize model training. Experiments on three well-known benchmark datasets for video classification show that our proposed truNcatE-split-contrAsT (NEAT) significantly outperforms the existing baselines. By reducing the dimension to 10% of it, our method achieves over 0.4 noise detection F1-score and 5% classification accuracy improvement on Mini-Kinetics dataset under severe noise (symmetric-80%). Thanks to Noise Contrastive Learning, the average classification accuracy improvement on Mini-Kinetics and Sth-Sth-V1 is over 1.6%.

The traditional model upgrading paradigm for retrieval requires recomputing all gallery embeddings before deploying the new model (dubbed as "backfilling"), which is quite expensive and time-consuming considering billions of instances in industrial applications. BCT presents the first step towards backward-compatible model upgrades to get rid of backfilling. It is workable but leaves the new model in a dilemma between new feature discriminativeness and new-to-old compatibility due to the undifferentiated compatibility constraints. In this work, we propose Darwinian Model Upgrades (DMU), which disentangle the inheritance and variation in the model evolving with selective backward compatibility and forward adaptation, respectively. The old-to-new heritable knowledge is measured by old feature discriminativeness, and the gallery features, especially those of poor quality, are evolved in a lightweight manner to become more adaptive in the new latent space. We demonstrate the superiority of DMU through comprehensive experiments on large-scale landmark retrieval and face recognition benchmarks. DMU effectively alleviates the new-to-new degradation at the same time improving new-to-old compatibility, rendering a more proper model upgrading paradigm in large-scale retrieval systems.Code: https://github.com/TencentARC/OpenCompatible.

It has been commonly observed that a teacher model with superior performance does not necessarily result in a stronger student, highlighting a discrepancy between current teacher training practices and effective knowledge transfer.In order to enhance the guidance of the teacher training process, we introduce the concept of distillation influence to determine the impact of distillation from each training sample on the student's generalization ability.In this paper, we propose Learning Good Teacher Matters (LGTM), an efficient training technique for incorporating distillation influence into the teacher's learning process.By prioritizing samples that are likely to enhance the student's generalization ability, our LGTM outperforms 10 common knowledge distillation baselines on 6 text classification tasks in the GLUE benchmark.1

High-fidelity facial avatar reconstruction from a monocular video is a significant research problem in computer graphics and computer vision. Recently, Neural Radiance Field (NeRF) has shown impressive novel view rendering results and has been considered for facial avatar reconstruction. However, the complex facial dynamics and missing 3D information in monocular videos raise significant challenges for faithful facial reconstruction. In this work, we propose a new method for NeRF-based facial avatar reconstruction that utilizes 3D-aware generative prior. Different from existing works that depend on a conditional deformation field for dynamic modeling, we propose to learn a personalized generative prior, which is formulated as a local and low dimensional subspace in the latent space of 3D-GAN. We propose an efficient method to construct the personalized generative prior based on a small set of facial images of a given individual. After learning, it allows for photo-realistic rendering with novel views, and the face reenactment can be realized by performing navigation in the latent space. Our proposed method is applicable for different driven signals, including RGB images, 3DMM coefficients, and audio. Compared with existing works, we obtain superior novel view synthesis results and faithfully face reenactment performance. The code is available here https://github.com/bbaaii/HFA-GP.

The goal of data-free meta-learning is to learn useful prior knowledge from a collection of pre-trained models without accessing their training data. However, existing works only solve the problem in parameter space, which (i) ignore the fruitful data knowledge contained in the pretrained models; (ii) can not scale to large-scale pre-trained models; (iii) can only meta-learn pre-trained models with the same network architecture. To address those issues, we propose a unified framework, dubbed PURER, which contains: (1) ePisode cUrriculum inveRsion (ECI) during data-free meta training; and (2) invErsion calibRation following inner loop (ICFIL) during meta testing. During meta training, we propose ECI to perform pseudo episode training for learning to adapt fast to new unseen tasks. Specifically, we progressively synthesize a sequence of pseudo episodes by distilling the training data from each pre-trained model. The ECI adaptively increases the difficulty level of pseudo episodes according to the real-time feedback of the meta model. We formulate the optimization process of meta training with ECI as an adversarial form in an end-to-end manner. During meta testing, we further propose a simple plug-and-play supplement—ICFIL—only used during meta testing to narrow the gap between meta training and meta testing task distribution. Extensive experiments in various real-world scenarios show the superior performance of ours.

Despite the remarkable success of foundation models, their task-specific fine-tuning paradigm makes them inconsistent with the goal of general perception modeling. The key to eliminating this inconsistency is to use generalist models for general task modeling. However, existing attempts at generalist models are inadequate in both versatility and performance. In this paper, we propose Uni-Perceiver v2, which is the first generalist model capable of handling major large-scale vision and vision-language tasks with competitive performance. Specifically, images are encoded as general region proposals, while texts are encoded via a Transformer-based language model. The encoded representations are transformed by a task-agnostic decoder. Different tasks are formulated as a unified maximum likelihood estimation problem. We further propose an effective optimization technique named Task-Balanced Gradient Normalization to ensure stable multi-task learning with an unmixed sampling strategy, which is helpful for tasks requiring large batch-size training. After being jointly trained on various tasks, Uni-Perceiver v2 is capable of directly handling downstream tasks without any task-specific adaptation. Results show that Uni-Perceiver v2 outperforms all existing generalist models in both versatility and performance. Meanwhile, compared with the commonly-recognized strong baselines that require tasks-specific fine-tuning, Uni-Perceiver v2 achieves competitive performance on a broad range of vision and vision-language tasks.

The real-world data tends to be heavily imbalanced and severely skew the data-driven deep neural networks, which makes Long-Tailed Recognition (LTR) a massive challenging task. Existing LTR methods seldom train Vision Transformers (ViTs) with Long-Tailed (LT) data, while the off-the-shelf pretrain weight of ViTs always leads to unfair comparisons. In this paper, we systematically investigate the ViTs' performance in LTR and propose LiVT to train ViTs from scratch only with LT data. With the observation that ViTs suffer more severe LTR problems, we conduct Masked Generative Pretraining (MGP) to learn generalized features. With ample and solid evidence, we show that MGP is more robust than supervised manners. Although Binary Cross Entropy (BCE) loss performs well with ViTs, it struggles on the LTR tasks. We further propose the balanced BCE to ameliorate it with strong theoretical groundings. Specially, we derive the unbiased extension of Sigmoid and compensate extra logit margins for deploying it. Our Bal-BCE contributes to the quick convergence of ViTs in just a few epochs. Extensive experiments demonstrate that with MGP and Bal-BCE, LiVT successfully trains ViTs well without any additional data and outperforms comparable state-of-the-art methods significantly, e.g., our ViT-B achieves 81.0% Top-1 accuracy in iNaturalist 2018 without bells and whistles. Code is available at https://github.com/XuZhengzhuo/LiVT.

Frequency-based methods have recently received much attention due to their impressive restoration of detail and structure in single image super-resolution (SISR). However, most of these methods mainly use frequency information as auxiliary means but ignore exploring the correlations and pixel distribution differences among various frequencies. To address the limitations, we propose a novel Frequency Reciprocal Action and Fusion Network (FRAF) that explores various frequency correlations and differences. Specifically, we design a Frequency Reciprocal Action (FRA) module, which safely enhances valid spatial information and decreases un-necessary repetition by reciprocal action among various spatial frequencies, to generate refined high- and low-frequency features. These refined frequency features are then progressively to guide the details and structure recovery, respectively. Furthermore, we develop a Detail and Structure Fusion (DSF) module to adaptively select, enhance and fuse the features to output the final HR image. This way ensures the final image is a high-quality product with rich details and a clear structure. Experimental results demonstrate that our method achieves superior performance over state-of-the-art (SOTA) approaches on both quantitative and qualitative evaluations.

In the real world, data tends to follow long-tailed distributions w.r.t. class or attribution, motivating the challenging Long-Tailed Recognition (LTR) problem. In this paper, we revisit recent LTR methods with promising Vision Transformers (ViT). We figure out that 1) ViT is hard to train with longtailed data. 2) ViT learns generalized features in an unsupervised manner, like mask generative training, either on longtailed or balanced datasets. Hence, we propose to adopt unsupervised learning to utilize long-tailed data. Furthermore, we propose the Predictive Distribution Calibration (PDC) as a novel metric for LTR, where the model tends to simply classify inputs into common classes. Our PDC can measure the model calibration of predictive preferences quantitatively. On this basis, we find many LTR approaches alleviate it slightly, despite the accuracy improvement. Extensive experiments on benchmark datasets validate that PDC reflects the model’s predictive preference precisely, which is consistent with the visualization.

3D Morphable Models (3DMMs) demonstrate great potential for reconstructing faithful and animatable 3D facial surfaces from a single image. The facial surface is influenced by the coarse shape, as well as the static detail (e.g., person-specific appearance) and dynamic detail (e.g., expression-driven wrinkles). Previous work struggles to decouple the static and dynamic details through image-level supervision, leading to reconstructions that are not realistic. In this paper, we aim at high-fidelity 3D face reconstruction and propose HiFace to explicitly model the static and dynamic details. Specifically, the static detail is modeled as the linear combination of a displacement basis, while the dynamic detail is modeled as the linear interpolation of two displacement maps with polarized expressions. We exploit several loss functions to jointly learn the coarse shape and fine details with both synthetic and real-world datasets, which enable HiFace to reconstruct high-fidelity 3D shapes with animatable details. Extensive quantitative and qualitative experiments demonstrate that HiFace presents state-of-the-art reconstruction quality and faithfully recovers both the static and dynamic details. Our project page: https://project-hiface.github.io.

Knowledge Distillation (KD) uses the teacher’s logits as soft labels to guide the student, while self-KD does not need a real teacher to require the soft labels. This work unifies the formulations of the two tasks by decomposing and reorganizing the generic KD loss into a Normalized KD (NKD) loss and customized soft labels for both target class (image’s category) and non-target classes named Universal Self-KD (USKD). We decompose the KD loss and find the non-target loss from it forces the student’s non-target logits to match the teacher’s, but the sum of the two non-target logits is different, preventing them from being identical. NKD normalizes the non-target logits to equalize their sum. It can be generally used for KD and self-KD to better use the soft labels for distillation. USKD generates customized soft labels for both target and non-target classes without a teacher. It smooths the target logit of the student as the soft target label and uses the rank of the intermediate feature to generate the soft non-target labels with Zipf’s law. For KD with teachers, NKD achieves state-of-the-art performance on CIFAR-100 and ImageNet, boosting the ImageNet Top-1 accuracy of Res-18 from 69.90% to 71.96% with a Res-34 teacher. For self-KD without teachers, USKD is the first method that can be effectively applied to both CNN and ViT models with negligible additional time and memory cost, resulting in new state-of-the-art results, such as 1.17% and 0.55% accuracy gains on ImageNet for MobileNet and DeiT-Tiny, respectively. Code is available at https://github.com/yzd-v/cls_KD.

3D GAN inversion aims to achieve high reconstruction fidelity and reasonable 3D geometry simultaneously from a single image input. However, existing 3D GAN inversion methods rely on time-consuming optimization for each individual case. In this work, we introduce a novel encoder-based inversion framework based on EG3D, one of the most widely-used 3D GAN models. We leverage the inherent properties of EG3D’s latent space to design a discriminator and a background depth regularization. This enables us to train a geometry-aware encoder capable of converting the input image into corresponding latent code. Additionally, we explore the feature space of EG3D and develop an adaptive refinement stage that improves the representation ability of features in EG3D to enhance the recovery of fine-grained textural details. Finally, we propose an occlusion-aware fusion operation to prevent distortion in unobserved regions. Our method achieves impressive results comparable to optimization-based methods while operating up to 500 times faster. Our framework is well-suited for applications such as semantic editing.

Accurately reconstructing 3D faces from monocular images and videos is crucial for various applications, such as digital avatar creation. However, the current deep learning-based methods face significant challenges in achieving accurate reconstruction with disentangled facial parameters and ensuring temporal stability in single-frame methods for 3D face tracking on video data. In this paper, we propose TokenFace, a transformer-based monocular 3D face reconstruction model. TokenFace uses separate tokens for different facial components to capture information about different facial parameters and employs temporal transformers to capture temporal information from video data. This design can naturally disentangle different facial components and is flexible to both 2D and 3D training data. Trained on hybrid 2D and 3D data, our model shows its power in accurately reconstructing faces from images and producing stable results for video data. Experimental results on popular benchmarks NoWand Stirling demonstrate that TokenFace achieves state-of-the-art performance, outperforming existing methods on all metrics by a large margin.

Real-world large-scale datasets are both noisily labeled and class-imbalanced. The issues seriously hurt the generalization of trained models. It is hence significant to address the simultaneous incorrect labeling and class-imbalance, i.e., the problem of learning with noisy labels on long-tailed data. Previous works develop several methods for the problem. However, they always rely on strong assumptions that are invalid or hard to be checked in practice. In this paper, to handle the problem and address the limitations of prior works, we propose a representation calibration method RCAL. Specifically, RCAL works with the representations extracted by unsupervised contrastive learning. We assume that without incorrect labeling and class imbalance, the representations of instances in each class conform to a multivariate Gaussian distribution, which is much milder and easier to be checked. Based on the assumption, we recover underlying representation distributions from polluted ones resulting from mislabeled and class-imbalanced data. Additional data points are then sampled from the recovered distributions to help generalization. Moreover, during classifier training, representation learning takes advantage of representation robustness brought by contrastive learning, which further improves the classifier performance. We derive theoretical results to discuss the effectiveness of our representation calibration. Experiments on multiple benchmarks justify our claims and confirm the superiority of the proposed method.

We study how to represent a video with implicit neural representations (INRs). Classical INRs methods generally utilize MLPs to map input coordinates to output pixels. While some recent works have tried to directly reconstruct the whole image with CNNs. However, we argue that both the above pixel-wise and image-wise strategies are not favorable to video data. Instead, we propose a patch-wise solution, PS-NeRV, which represents videos as a function of patches and the corresponding patch coordinate. It naturally inherits the advantages of image-wise methods, and achieves excellent reconstruction performance with fast decoding speed. The whole method includes conventional modules, like positional embedding, MLPs and CNNs. We also introduce AdaIN to enhance intermediate features. Extensive experiments have demonstrated its effectiveness in several video-related tasks, such as video compression and video inpainting.

We address the problem of photorealistic 3D face avatar synthesis from sparse images. Existing Parametric models for face avatar reconstruction struggle to generate details that originate from inputs. Meanwhile, although current NeRF-based avatar methods provide promising results for novel view synthesis, they fail to generalize well for unseen expressions. We improve from NeRF and propose a novel framework that, by leveraging the parametric 3DMM models, can reconstruct a high-fidelity drivable face avatar and successfully handle the unseen expressions. At the core of our implementation are structured displacement feature and semantic-aware learning module. Our structured displacement feature will introduce the motion prior as an additional constraints and help perform better for unseen expressions, by constructing displacement volume. Besides, the semantic-aware learning incorporates multi-level prior, e.g., semantic embedding, learnable latent code, to lift the performance to a higher level. Thorough experiments have been doen both quantitatively and qualitatively to demonstrate the design of our framework, and our method achieves much better results than the current state-of-the-arts.

Data-free meta-learning (DFML) aims to enable efficient learning of new tasks by meta-learning from a collection of pre-trained models without access to the training data. Existing DFML work can only meta-learn from (i) white-box and (ii) small-scale pre-trained models (iii) with the same architecture, neglecting the more practical setting where the users only have inference access to the APIs with arbitrary model architectures and model scale inside. To solve this issue, we propose a Bi-level Data-free Meta Knowledge Distillation (BiDf-MKD) framework to transfer more general meta knowledge from a collection of black-box APIs to one single meta model. Specifically, by just querying APIs, we inverse each API to recover its training data via a zero-order gradient estimator and then perform meta-learning via a novel bi-level meta knowledge distillation structure, in which we design a boundary query set recovery technique to recover a more informative query set near the decision boundary. In addition, to encourage better generalization within the setting of limited API budgets, we propose task memory replay to diversify the underlying task distribution by covering more interpolated tasks. Extensive experiments in various real-world scenarios show the superior performance of our BiDf-MKD framework.

Real-world data contains a vast amount of multimodal information, among which vision and language are the two most representative modalities. Moreover, increasingly heavier models, e.g., Transformers, have attracted the attention of researchers to model compression. However, how to compress multimodal models, especially vison-language Transformers, is still under-explored. This paper proposes the Unified and Progressive Pruning (UPop) as a universal vison-language Transformer compression framework, which incorporates 1) unifiedly searching multimodal subnets in a continuous optimization space from the original model, which enables automatic assignment of pruning ratios among compressible modalities and structures; 2) progressively searching and retraining the subnet, which maintains convergence between the search and retrain to attain higher compression ratios. Experiments on various tasks, datasets, and model architectures demonstrate the effectiveness and versatility of the proposed UPop framework. The code is available at https://github.com/sdc17/UPop.

Visual place recognition (VPR) is one of the research hotspots in robotics, which uses visual information to locate robots. Recently, the hierarchical two-stage VPR methods have become popular in this field due to the trade-off between accuracy and efficiency. These methods retrieve the top-k candidate images using the global features in the first stage, then re-rank the candidates by matching the local features in the second stage. However, they usually require additional al-gorithms (e.g. RANSAC) for geometric consistency verification in re-ranking, which is time-consuming. Here we propose a Dynamically Aligning Local Features (DALF) algorithm to align the local features under spatial constraints. It is significantly more efficient than the methods that need geometric consistency verification. We present a unified network capable of extracting global features for retrieving candidates via an aggregation module and aligning local features for re-ranking via the DALF alignment module. We call this network AANet. Meanwhile, many works use the simplest positive samples in triplet for weakly supervised training, which limits the ability of the network to recognize harder positive pairs. To address this issue, we propose a Semi-hard Positive Sample Mining (ShPSM) strategy to select appropriate hard positive images for training more robust VPR networks. Extensive experiments on four benchmark VPR datasets show that the proposed AANet can outperform several state-of-the-art methods with less time consumption. The code is released at https://github.com/Lu-Feng/AANet.

Recently, techniques utilizing frequency-based methods have gained significant attention, as they exhibit exceptional restoration capabilities for detail and structure in video super-resolution tasks. However, most of these frequency-based methods mainly have three major limitations: 1) insufficient exploration of object motion information, 2) inadequate enhancement for high-fidelity regions, and 3) loss of spatial information during convolution. In this paper, we propose a novel network, Directional Frequency Video Super-Resolution (DFVSR), to address these limitations. Specifically, we reconsider object motion from a new perspective and propose Directional Frequency Representation (DFR), which not only borrows the property of frequency representation of detail and structure information but also contains the direction information of the object motion that is extremely significant in videos. Based on this representation, we propose a Directional Frequency-Enhanced Alignment (DFEA) to use double enhancements of task-related information for ensuring the retention of high-fidelity frequency regions to generate the high-quality alignment feature. Furthermore, we design a novel Asymmetrical U-shaped network architecture to progressively fuse these alignment features and output the final output. This architecture enables the intercommunication of the same level of resolution in the encoder and decoder to achieve the supplement of spatial information. Powered by the above designs, our method achieves superior performance over state-of-the-art models on both quantitative and qualitative evaluations.

Most of these frequency-based deblurring methods mainly have two major limitations: (1) insufficient exploitation of frequency information, (2) inadequate preservation of frequency information. In this paper, we propose a novel Efficient Frequency Exploitation and Preservation Network (EFEP) to address these limitations. Firstly, we propose a novel Frequency-Balanced Exploitation Encoder (FBE-Encoder) to sufficiently exploit frequency information. We insert a novel Frequency-Balanced Navigator (FBN) module in the encoder, which establishes a dynamic balance that adaptively explores and integrates the correlations between frequency features and other features presented in the network. And it also can highlight the most important regions in frequency features. Secondly, considering the limitation that frequency information is inevitably lost in deep network architectures, we present an Enhanced Selective Frequency Decoder (ESF-Decoder) that not only effectively reduces spatial information redundancy, but also fully explores the different importance of various frequency information to ensure the supplement of valid spatial information and weaken the invalid information. Thirdly, each encoder/decoder block of the EFEP consists of multiple Contrastive Residual Blocks (CRBs), which are designed to explicitly compute and incorporate feature distinctions. Powered by the above designs, our EFEP outperforms state-of-the-art models on both quantitative and qualitative evaluations.

Deep Neural Networks suffer significant performance degeneration when noisy labels corrupt latent data representations. Previous work has attempted to alleviate this problem by exploiting contrastive learning, the pair building of which is critical. However, existing methods either conduct sample-level processes and then use the resultant subset to construct pairs or directly perform pair-level selecting using a fixed threshold, both leading to sub-optimal pairing and subsequent representation learning. To address this issue, we propose a novel adaptive contrastive learning method (ACL) working at the pair level to select contrastive pairs adaptively. Specifically, we consider the model's learning status to adjust the confidence threshold in a self-adaptive manner instead of fixing it. Then, towards the ineffectiveness of the thresholding method on unconfident pairs, we automatically apply instance-specific temperature to boost the confidence of accurately-predicted samples and their pairs. We further introduce temporal cross-ensembling to handle the impact of noisy labels on model predictions. As a result, diverse pairs are correctly selected for contrastive learning to induce discriminative representations robust to various types of label noise. Extensive experimental results on several standard benchmarks and real-world datasets indicate the superiority of ACL, especially in extremely noisy scenarios.

Graph neural networks, which typically exchange information between local neighbors, often struggle to capture long-range interactions (LRIs) within the graph. Building a graph hierarchy via graph pooling methods is a promising approach to address this challenge; however, hierarchical information propagation cannot entirely take over the role of local information aggregation. To balance locality and hierarchy, we integrate the local and hierarchical structures, represented by intra- and inter-graphs respectively, of a multi-scale graph hierarchy into a single mega graph. Our proposed MeGraph model consists of multiple layers alternating between local and hierarchical information aggregation on the mega graph. Each layer first performs local-aware message-passing on graphs of varied scales via the intra-graph edges, then fuses information across the entire hierarchy along the bidirectional pathways formed by inter-graph edges. By repeating this fusion process, local and hierarchical information could intertwine and complement each other. To evaluate our model, we establish a new Graph Theory Benchmark designed to assess LRI capture ability, in which MeGraph demonstrates dominant performance. Furthermore, MeGraph exhibits superior or equivalent performance to state-of-the-art models on the Long Range Graph Benchmark. The experimental results on commonly adopted real-world datasets further demonstrate the broad applicability of MeGraph.

Goal-conditioned reinforcement learning (RL) with sparse rewards remains a challenging problem in deep RL. Hindsight Experience Replay (HER) has been demonstrated to be an effective solution, where HER replaces desired goals in failed experiences with practically achieved states. Existing approaches mainly focus on either exploration or exploitation to improve the performance of HER. From a joint perspective, exploiting specific past experiences can also implicitly drive exploration. Therefore, we concentrate on prioritizing both original and relabeled samples for efficient goal-conditioned RL. To achieve this, we propose a novel value consistency prioritization (VCP) method, where the priority of samples is determined by the consistency of ensemble Q-values. This distinguishes the VCP method with most existing prioritization approaches which prioritizes samples based on the uncertainty of ensemble Q-values. Through extensive experiments, we demonstrate that VCP achieves significantly higher sample efficiency than existing algorithms on a range of challenging goal-conditioned manipulation tasks. We also visualize how VCP prioritizes good experiences to enhance policy learning.

Few-shot classification is a challenging task of computer vision and is critical to the data-sparse scenario like rare disease diagnosis. Feature augmentation is a straightforward way to alleviate the data-sparse issue in few-shot classification. However, mimicking the original feature distribution from a small amount of data is challenging. Existing augmentation-based methods are task-agnostic: the augmented feature is not with optimal intra-class diversity and inter-class discriminability concerning a certain task. To address this drawback, we propose a novel Task-adaptive Feature Disentanglement and Hallucination framework, dubbed TaFDH. Concretely, we first perceive the task information to disentangle the original feature into two components: class-irrelevant and class-specific features. Then more class-irrelevant features are decoded from a learned variational distribution, fused with the class-specific feature to get the augmented features. Finally, a generalized prior distribution over a quadratic classifier is meta-learned, which can be fast adapted to the class-specific posterior, thus further alleviating the inadequacy and uncertainty of feature hallucination via the nature of Bayesian inference. In this way, we construct a more discriminable embedding space with reasonable intra-class diversity instead of simply restoring the original embedding space, which can lead to a more precise decision boundary. We obtain the augmented features equipped with enhanced inter-class discriminability by highlighting the most discriminable part while boosting the intra-class diversity by fusing with the diverse generated class-irrelevant parts. Experiments on five multi-grained few-shot classification datasets demonstrate the superiority of our method.

It is desirable to develop efficient image rescaling methods to transmit digital images with different resolutions between devices and assure visual quality. In image downscaling, the inevitable loss of high-frequency information makes the reverse upscaling highly ill-posed. Recent approaches focus on joint learning of image downscaling and upscaling (e.g., rescaling). However, existing methods still fail to recover satisfactory high-frequency signals when upscaling. To solve it, we propose high-frequency flow (HfFlow), which learns the distribution of high-frequency signals during rescaling. HfFlow is an overall invertible framework with a conditional flow on the high-frequency space to compensate for the information lost during downscaling. To facilitate finding the optimal upscaling solution, we introduce a reference low-resolution (LR) manifold and propose a cross-entropy Gaussian loss (CGloss) to force the downscaled manifold closer to the reference LR manifold and simultaneously fulfill recovering missing details. HfFlow can be generalized to other scale transformation tasks such as image colorization with its excellent rescaling capacity. Qualitative and quantitative experimental evaluations demonstrate that HfFlow restores rich high-frequency details and outperforms state-of-the-art rescaling methods in PSNR, SSIM, and perceptual quality metrics.

Scene text detection is still a challenging task, as there may be extremely small or low-resolution strokes and close or arbitrary-shaped texts. In this paper, StrokeNet proposes to effectively detect the texts by capturing the fine-grained strokes and inferring structural relations between the hierarchical representations of each text area in the graph-based network. Different from existing approaches that represent the text area by a series of points or rectangular boxes, we directly localize the strokes of each text instance. We introduce Stroke Assisted Prediction Network (SAPN), which performs hierarchical representation learning of text areas, effectively capturing extremely small or low-resolution texts. We extract a series of text- and stroke-level rectangular boxes on the predicted text areas, which are treated as graph nodes and grouped to form the corresponding local graphs. Hierarchical Relation Graph Network (HRGN) then performs relational reasoning and predicts the likelihood of linkages among graph nodes of different levels. It efficiently splits the close text instances and grouping node classification results into the arbitrary-shaped text area. We introduce a novel dataset with stroke-level annotations, namelySynthStroke, for offline pre-training of widespread text detectors. Experiments on benchmarks verify the State-of-the-Art performance of our method.

Deep hashing has shown promising performance in large-scale image retrieval. The hashing process utilizes Deep Neural Networks (DNNs) to embed images into compact continuous latent codes, then map them into binary codes by hashing function for efficient retrieval. Recent approaches perform metric loss and quantization loss to supervise the two procedures that cluster samples with the same categories and alleviate semantic information loss after binarization in the end-to-end training framework. However, we observe the incompatible conflict that the optimal cluster positions are not identical to the ideal hash positions because of the different objectives of the two loss terms, which lead to severe ambiguity and error-hashing after the binarization process. To address the problem, we borrow the Theory of Minimum-Distance Bounds for Binary Linear Codes to design the inflection point that depends on the hash bit length and category numbers and thereby propose Hashing-guided Hinge Function (HHF) to explicitly enforce the termination of metric loss to prevent the negative pairs unlimited alienated. Such modification is proven effective and essential for training, which contributes to proper intra- and inter-distances for clusters and better hash positions for accurate image retrieval simultaneously. Extensive experiments in CIFAR-10, CIFAR-100, ImageNet, and MS-COCO justify that HHF consistently outperforms existing techniques and is robust and flexible to transplant into other methods. Code is available athttps://github.com/JerryXu0129/HHF.

Weakly supervised instance segmentation with image-level class supervision is a challenging task as it associates the highest-level instances to the lowest-level appearance. Previous approaches for the task utilize classification networks to obtain rough discriminative parts as seed regions and use distance as a metric to cluster pixels of the same instances. Unlike previous approaches, we provide a novel self-supervised joint learning framework as the basic network and consider the clustering problem as calculating the probability that pixels belong to each instance. To this end, we propose our self-supervised joint learning two-stream network (SJLT Net) to finish this task. In the first stream, we leverage a joint learning framework to implement image-level supervised semantic segmentation with self-supervised saliency detection. In the second stream, we propose a Center Detection Network to detect different instances’ centers with the gaussian loss function to cluster instances pixels. Besides, an integration module is utilized to combine information of both streams and get precise pseudo instances labels. Our approach generates pseudo instance segmentation labels of training images, which are used to train a fully supervised model. Our model achieves excellent performance on the PASCAL VOC 2012 dataset, surpassing the best baseline trained with the same labels by 4.6$\%$$AP^r_{50}$on the train set and 2.6$\%$$AP^r_{50}$on the validation set.

Matching-based methods, especially those based on space-time memory, are significantly ahead of other solutions in semi-supervised video object segmentation (VOS). However, continuously growing and redundant template features lead to an inefficient inference. To alleviate this, we propose a novel Sequential Weighted Expectation-Maximization (SWEM) network to greatly reduce the redundancy of memory features. Different from the previous methods which only detect feature redundancy between frames, SWEM merges both intra-frame and inter-frame similar features by leveraging the sequential weighted EM algorithm. Further, adaptive weights for frame features endow SWEM with the flexibility to represent hard samples, improving the discrimination of templates. Besides, the proposed method maintains a fixed number of template features in memory, which ensures the stable inference complexity of the VOS system. Extensive experiments on commonly used DAVIS and YouTube-VOS datasets verify the high efficiency (36 FPS) and high performance (84.3% J&F on DAVIS 2017 validation dataset) of SWEM.

Knowledge distillation has been applied to image classification successfully. However, object detection is much more sophisticated and most knowledge distillation methods have failed on it. In this paper, we point out that in object detection, the features of the teacher and student vary greatly in different areas, especially in the foreground and background. If we distill them equally, the uneven differences between feature maps will negatively affect the distillation. Thus, we propose Focal and Global Distillation (FGD). Focal distillation separates the foreground and background, forcing the student to focus on the teacher's critical pixels and channels. Global distillation rebuilds the relation between different pixels and transfers it from teachers to students, compensating for missing global information in focal distillation. As our method only needs to calculate the loss on the feature map, FGD can be applied to various detectors. We experiment on various detectors with different backbones and the results show that the student detector achieves excellent mAP improvement. For example, ResNet-50 based RetinaNet, Faster RCNN, RepPoints and Mask RCNN with our distillation method achieve 40.7%, 42.0%, 42.0% and 42.1% mAP on COCO2017, which are 3.3, 3.6, 3.4 and 2.9 higher than the baseline, respectively. Our codes are available at https://github.com/yzd-v/FGD.

Spatiotemporal predictive learning (ST-PL) aims at predicting the subsequent frames via limited observed sequences, and it has broad applications in the real world. However, learning representative spatiotemporal features for prediction is challenging. Moreover, chaotic uncertainty among consecutive frames exacerbates the difficulty in long-term prediction. This paper concentrates on improving prediction quality by enhancing the correspondence between the previous context and the current state. We carefully design Detail Context Block (DCB) to extract fine-grained details and improve the isolated correlation between upper context state and current input state. We integrate DCB with standard ConvLSTM and introduce Motion Details RNN (MoDeRNN) to capture fine-grained spatiotemporal features and improve the expression of latent states of RNNs to achieve significant quality. Experiments on Moving MNIST and Typhoon datasets demonstrate the effectiveness of the proposed method. MoDeRNN outperforms existing state-of-the-art techniques qualitatively and quantitatively with lower computation loads.

For medical image segmentation in a real scenario, the amount of accurate annotation data at the pixel level is typically small, which tends to cause an overfitting problem. This manuscript goes deep into the research of the Dropout algorithm, which is commonly used in neural networks to alleviate the overfitting problem. From the perspective of solving the co-adaptation problem, this manuscript explains the basic principles of the Dropout algorithm and discusses the existing limitations of its derivative methods. Furthermore, we propose a novel Heuristic Dropout algorithm to address these limitations. The proposed algorithm takes information entropy and variance as heuristic rules. It guides our algorithm to drop features suffering from co-adaptation problem more efficiently and thus can better alleviate the overfitting problem of small-scale medical image segmentation datasets. Experiments on medical image segmentation datasets and models show that the proposed algorithm significantly improves the performance of these models.

Spatiotemporal predictive learning (ST-PL) is a hotspot with numerous applications, such as object movement and mete-orological prediction. It aims at predicting the subsequent frames via observed sequences. However, inherent uncer-tainty among consecutive frames exacerbates the difficulty in long-term prediction. To tackle the increasing ambigu-ity during forecasting, we design CMS-LSTM to focus on context correlations and multi-scale spatiotemporal flow with details on fine-grained locals, containing two elaborate de-signed blocks: Context Embedding (CE) and Spatiotemporal Expression (SE) blocks. CE is designed for abundant context interactions, while SE focuses on multi-scale spatiotemporal expression in hidden states. The newly introduced blocks also facilitate other spatiotemporal models (e.g., PredRNN, SA-ConvLSTM) to produce representative implicit features for ST-PL and improve prediction quality. Qualitative and quanti-tative experiments demonstrate the effectiveness and flexibil-ity of our proposed method. With fewer params, CMS-LSTM outperforms state-of-the-art methods in numbers of metrics on two representative benchmarks and scenarios. Code is available at https://github.com/czh-98/CMS-LSTM.

Since single-modal controllable manipulation typically requires supervision of information from other modalities or cooperation with complex software and experts, this paper addresses the problem of cross-modal adaptive manipulation (CAM). The novel task performs cross-modal semantic alignment from mutual supervision and implements bidirectional exchange of attributes, relations, or objects in parallel, benefiting both modalities while significantly reducing manual effort. We introduce a robust solution for CAM, which includes two essential modules, namely Heterogeneous Representation Learning (HRL) and Cross-modal Relation Reasoning (CRR). The former is designed to perform representation learning for cross-modal semantic alignment on heterogeneous graph nodes. The latter is adopted to identify and exchange the focused attributes, relations, or objects in both modalities. Our method produces pleasing cross-modal outputs on CUB and Visual Genome.

Conventional model upgrades for visual search systems require offline refresh of gallery features by feeding gallery images into new models (dubbed as “backfill”), which is time-consuming and expensive, especially in large-scale applications. The task of backward-compatible representation learning is therefore introduced to support backfill-free model upgrades, where the new query features are interoperable with the old gallery features. Despite the success, previous works only investigated a close-set training scenario (i.e., the new training set shares the same classes as the old one), and are limited by more realistic and challenging open-set scenarios. To this end, we first introduce a new problem of universal backward-compatible representation learning, covering all possible data split in model upgrades. We further propose a simple yet effective method, dubbed as Universal Backward-Compatible Training (UniBCT) with a novel structural prototype refinement algorithm, to learn compatible representations in all kinds of model upgrading benchmarks in a unified manner. Comprehensive experiments on the large-scale face recognition datasets MS1Mv3 and IJB-C fully demonstrate the effectiveness of our method. Source code is available at https://github.com/TencentARC/OpenCompatible.

Image dehazing remains a challenging problem because it is hard to restore a clean scene from a severely degraded hazy image. However, existing learning-based dehazing methods mostly ignore the fact that the interference of haze to an image is mainly concentrated in the low-frequency components. If all image components are processed indiscriminately, it is difficult to achieve a good restoration and accurate details cannot be guaranteed. In order to process the hazy images hierarchically, we propose a low-frequency sub-band contrastive regularization (LSCR) in the wavelet domain to ensure that the components of the restored image mainly affected by haze are pulled closer to the clear image and pushed far away from the hazy image. In addition, a high-frequency sub-band loss is also introduced to make high-frequency components of the restored image consistent with the clear image. Our method can better restore the haze-free image and achieve more accurate and rich details. The extensive experiments on synthetic and real-world datasets verify that the proposed method outperforms previous approaches.

Most existing weakly supervised aspect detection algorithms utilize pre-trained language models as their backbone networks by constructing discriminative tasks with seed words. Once the number of seed words decreases, the performance of current models declines significantly. Recently, prompt tuning has been proposed to bridge the gap of objective forms in pre-training and fine-tuning, which is hopeful of alleviating the above challenge. However, directly applying the existing prompt-based methods to this task not only fails to effectively use large amounts of unlabeled data, but also may cause serious over-fitting problems. In this paper, we propose a lightweight teacher-student network (PTS) based on prompts to solve the above two problems. Concretely, the student network is a hybrid prompt-based classification model to detect aspects, which innovatively compounds hand-crafted prompts and auto-generated prompts. The teacher network comprehensively considers the representation of the sentence and the masked aspect token in the template to guide classification. To utilize unlabeled data and seed words intelligently, we train the teacher and student network alternately. Furthermore, in order to solve the problem that the uneven quality of training data obviously affects the iterative efficiency of PTS, we design a general dynamic data selection strategy to feed the most pertinent data into the current model. Experimental results show that even given the minimum seed words, PTS significantly outperforms previous state-of-the-art methods on three widely used benchmarks.

Unsupervised aspect detection (UAD) aims to identify the aspect categories mentioned in product reviews automatically. Existing unsupervised methods mainly focus on single-label prediction which cannot work well for the realistic application scene since a review usually contain multiple aspect categories. Recent attempts alleviate this issue by setting a threshold. However, the imbalance of the number of category-related words in review segments makes these methods difficult to find optimal thresholds to recall all categories, which is a common but neglected phenomenon. In this paper, we propose a novel unsupervised method termed Aspect-Category Experts (ACEs) to address this problem. Our goal is to train a set of aspect-category experts to encode the sentence in parallel, where experts and aspect categories correspond one-to-one. Experts in different aspects weight the embedding of representative words with aspect-specific attention to avoid the negative impact of accumulation. Besides, to enhance the complementarity between different experts to reduce inter-class feature entanglement, we construct a novel mutual exclusion loss (ME loss) to improve the aspect detection performance. Extensive experimental results on four datasets demonstrate that our proposed ACEs model outperforms the previous state-of-the-art methods.

This paper presents a novel video inpainting architecture named Deformed Vision Transformers (DeViT). We make three significant contributions to this task: First, we extended previous Transformers with patch alignment by introducing Deformed Patch-based Homography Estimator (DePtH), which enriches the patch-level feature alignments in key and query with additional offsets learned from patch pairs without additional supervision. DePtH enables our method to handle challenging scenes or agile motion with in-plane or out-of-plane deformation, which previous methods usually fail. Second, we introduce the Mask Pruning-based Patch Attention (MPPA) to improve the standard patch-wised feature matching by pruning out less essential features and considering the saliency map. MPPA enhances the matching accuracy between warped tokens with invalid pixels. Third, we introduce the Spatial-Temporal weighting Adaptor (STA) module to assign more accurate attention to spatial-temporal tokens under the guidance of the Deformation Factor learned from DePtH, especially for videos with agile motions. Experimental results demonstrate that our method outperforms previous state-of-the-art methods in quality and quantity and achieves a new state-of-the-art for video inpainting.

Image retrieval has become an increasingly appealing technique with broad multimedia application prospects, where deep hashing serves as the dominant branch towards low storage and efficient retrieval. In this paper, we carried out in-depth investigations on metric learning in deep hashing for establishing a powerful metric space in multi-label scenarios, where the pair loss suffers high computational overhead and converge difficulty, while the proxy loss is theoretically incapable of expressing the profound label dependencies and exhibits conflicts in the constructed hypersphere space. To address the problems, we propose a novel metric learning framework with Hybrid Proxy-Pair Loss (HyP$^2$ Loss) that constructs an expressive metric space with efficient training complexity w.r.t. the whole dataset. The proposed HyP$^2$ Loss focuses on optimizing the hypersphere space by learnable proxies and excavating data-to-data correlations of irrelevant pairs, which integrates sufficient data correspondence of pair-based methods and high-efficiency of proxy-based methods. Extensive experiments on four standard multi-label benchmarks justify the proposed method outperforms the state-of-the-art, is robust among different hash bits and achieves significant performance gains with a faster, more stable convergence speed. Our code is available at https://github.com/JerryXu0129/HyP2-Loss.

Instance-dependent label noise is realistic but rather challenging, where the label-corruption process depends on instances directly. It causes a severe distribution shift between the distributions of training and test data, which impairs the generalization of trained models. Prior works put great effort into tackling the issue. Unfortunately, these works always highly rely on strong assumptions or remain heuristic without theoretical guarantees. In this paper, to address the distribution shift in learning with instance-dependent label noise, a dynamic distribution-calibration strategy is adopted. Specifically, we hypothesize that, before training data are corrupted by label noise, each class conforms to a multivariate Gaussian distribution at the feature level. Label noise produces outliers to shift the Gaussian distribution. During training, to calibrate the shifted distribution, we propose two methods based on the mean and covariance of multivariate Gaussian distribution respectively. The mean-based method works in a recursive dimension-reduction manner for robust mean estimation, which is theoretically guaranteed to train a high-quality model against label noise. The covariance-based method works in a distribution disturbance manner, which is experimentally verified to improve the model robustness. We demonstrate the utility and effectiveness of our methods on datasets with synthetic label noise and real-world unknown noise.

Federated learning (FL) is a privacy-preserving distributed learning paradigm that enables collaborative training of machine learning models among multiple participants. However, despite recent progress, existing federated learning systems can still not handle heterogeneous models. For instance, candidate clients with heterogeneous models are inaccessible to the established federated system. And within the federated system, local models are forbidden to be updated to become heterogeneous models, even though the updated models work better.

Free-form text prompts allow users to describe their intentions during image manipulation conveniently. Based on the visual latent space of StyleGAN[21] and text embedding space of CLIP[34], studies focus on how to map these two latent spaces for text-driven attribute manipulations. Currently, the latent mapping between these two spaces is empirically designed and confines that each manipulation model can only handle one fixed text prompt. In this paper, we propose a method named Free-Form CLIP (FFCLIP), aiming to establish an automatic latent mapping so that one manipulation model handles free-form text prompts. Our FFCLIP has a cross-modality semantic modulation module containing semantic alignment and injection. The semantic alignment performs the automatic latent mapping via linear transformations with a cross attention mechanism. After alignment, we inject semantics from text prompt embeddings to the StyleGAN latent space. For one type of image (e.g., human portrait'), one FFCLIP model can be learned to handle free-form text prompts. Meanwhile, we observe that although each training text prompt only contains a single semantic meaning, FFCLIP can leverage text prompts with multiple semantic meanings for image manipulation. In the experiments, we evaluate FFCLIP on three types of images (i.e.,human portraits', cars', andchurches'). Both visual and numerical results show that FFCLIP effectively produces semantically accurate and visually realistic images. Project page: https://github.com/KumapowerLIU/FFCLIP.

Real-Time Bidding (RTB) is an important mechanism in online search advertising involved with the interaction among platform, advertisers, and customers, where the proper bid for each page views plays a crucial role for good marketing results. Enhanced Cost Per Click (ECPC) bidding is one of typical bidding strategies in the RTB mechanism where the advertisers will raise bid prices for traffic with high conversion rates under the premise of total advertisers' cost control in different time periods of the day. The bid ratio in ECPC as a correction coefficient is used to control the advertising cost of advertisers. However, the optimal bid ratio is hard to be derived due to the complexity and volatility of the auction environment. To address the challenges, we formulate the hourly-aggregated level of ECPC bidding as a Markov Decision Process with innovative state and reward function according to business. Besides, quite different from prior work that has to construct an auction model-based environment to solve the MDP problem, we propose a novel framework based on model-free reinforcement learning that does not rely on the agent's explorations in the environment. We furthermore design an adaptive action choosing strategy which adjusts the bidding behavior dynamically and improves the performance of categories/groups's consumption qualified rate. Experimental results on real dataset demonstrate the effectiveness of our framework from the perspective of consistency, stability and online effect.

In this paper, we propose a new task of deep interactive video inpainting and an application for users to interact with machines. To our best knowledge, this is the first deep learning-based interactive video inpainting framework that only uses a free form of user input as guidance (i.e. scribbles) instead of mask annotations, which has academic, entertainment, and commercial value.

Vector Quantized Variational AutoEncoder (VQ-VAE) models realize fast image generation by encoding and quantifying the raw input in the single-level or hierarchical compressed latent space. However, the learned representations are not expert in capturing complex relations existed, while one usually adopts domain-specific autoregressive models to fit a prior distribution for two stages of learning. In this work, we propose VQMG, a novel and unified framework for multi-hops relational reasoning and explicit representation learning. By introducing Multi-hops Graph Convolution Networks (MGCN), complicated relations from hierarchical latent space are effectively captured by Inner Graph, while the fitting of autoregressive prior are performed coherently by Outer Graph to promote the performance. Experiments on multimedia tasks including Point cloud segementation, Stroke-level text detection and Image generation verify the efficiency and applicability of our approach.

Real-world data universally confronts a severe class-imbalance problem and exhibits a long-tailed distribution, i.e., most labels are associated with limited instances. The naïve models supervised by such datasets would prefer dominant labels, encounter a serious generalization challenge and become poorly calibrated. We propose two novel methods from the prior perspective to alleviate this dilemma. First, we deduce a balance-oriented data augmentation named Uniform Mixup (UniMix) to promote mixup in long-tailed scenarios, which adopts advanced mixing factor and sampler in favor of the minority. Second, motivated by the Bayesian theory, we figure out the Bayes Bias (Bayias), an inherent bias caused by the inconsistency of prior, and compensate it as a modification on standard cross-entropy loss. We further prove that both the proposed methods ensure the classification calibration theoretically and empirically. Extensive experiments verify that our strategies contribute to a better-calibrated model, and their combination achieves state-of-the-art performance on CIFAR-LT, ImageNet-LT, and iNaturalist 2018.

For further progress in video object segmentation (VOS), larger, more diverse, and more challenging datasets will be necessary. However, densely labeling every frame with pixel masks does not scale to large datasets. We use a deep convolutional network to automatically create pseudo-labels on a pixel level from much cheaper bounding box annotations and investigate how far such pseudo-labels can carry us for training state-of-the-art VOS approaches. A very encouraging result of our study is that adding a manually annotated mask in only a single video frame for each object is sufficient to generate pseudo-labels which can be used to train a VOS method to reach almost the same performance level as when training with fully segmented videos. We use this workflow to create pixel pseudo-labels for the training set of the challenging tracking dataset TAO, and we manually annotate a subset of the validation set. Together, we obtain the new TAO-VOS benchmark, which we make publicly available at www.vision.rwth-aachen.de/page/taovos. While the performance of state-of-the-art methods on existing datasets starts to saturate, TAO-VOS remains very challenging for current algorithms and reveals their shortcomings.

In transfer learning, how to effectively transfer useful information from the source domain to the target domain is crucial. In this paper, we propose a novel transfer learning method for image classification, named manifold transfer learning via discriminant regression analysis (MTL-DRA), to transfer the local geometry structure information from the source domain to the target domain and ensure that the transform matrix is robust or sparse so that samples from different domains can be well combined. In MTL-DRA, we encode discriminant information of the source domain to the target domain by introducing between- and within-class graphs to preserve within-class similarity and reduce between-class similarity. With different norms as constraints, MTL-DRA overcomes the disturbance of noise and avoids negative transfer learning. To improve the robustness of MTL-DRA, we encode a nuclear norm constraint and propose robust MTL-DRA (RMTL-DRA). We analyzed the convergence and complexity of the two proposed methods. To verify the performance of the proposed methods, we conducted extensive experiments on five public image benchmarks. The experimental results show that the proposed methods outperform state-of-the-art transfer learning methods.

Spatiotemporal prediction is challenging due to the complex dynamic motion and appearance changes. Existing work concentrates on embedding additional cells into the standard ConvLSTM to memorize spatial appearances during the prediction. These models always rely on the convolution layers to capture the spatial dependence, which are local and inefficient. However, long-range spatial dependencies are significant for spatial applications. To extract spatial features with both global and local dependencies, we introduce the self-attention mechanism into ConvLSTM. Specifically, a novel self-attention memory (SAM) is proposed to memorize features with long-range dependencies in terms of spatial and temporal domains. Based on the self-attention, SAM can produce features by aggregating features across all positions of both the input itself and memory features with pair-wise similarity scores. Moreover, the additional memory is updated by a gating mechanism on aggregated features and an established highway with the memory of the previous time step. Therefore, through SAM, we can extract features with long-range spatiotemporal dependencies. Furthermore, we embed the SAM into a standard ConvLSTM to construct a self-attention ConvLSTM (SA-ConvLSTM) for the spatiotemporal prediction. In experiments, we apply the SA-ConvLSTM to perform frame prediction on the MovingMNIST and KTH datasets and traffic flow prediction on the TexiBJ dataset. Our SA-ConvLSTM achieves state-of-the-art results on both datasets with fewer parameters and higher time efficiency than previous state-of-the-art method.

We study the commonsense inference task that aims to reason and generate the causes and effects of a given event. Existing neural methods focus more on understanding and representing the event itself, but pay little attention to the relations between different commonsense dimensions (e.g. causes or effects) of the event, making the generated results logically inconsistent and unreasonable. To alleviate this issue, we propose Chain Transformer, a logic enhanced commonsense inference model that combines both direct and indirect inferences to construct a logical chain so as to reason in a more logically consistent way. First, we apply a self-attention based encoder to represent and encode the given event. Then a chain of decoders is implemented to reason and generate for different dimensions following the logical chain, where an attention module is designed to link different decoders and to make each decoder attend to the previous reasoned inferences. Experiments on two real-world datasets show that Chain Transformer outperforms previous methods on both automatic and human evaluation, and demonstrate that Chain Transformer can generate more reasonable and logically consistent inference results.

Recent image captioning models have made much progress for exploring the multi-modal interaction, such as attention mechanisms.Though these mechanisms can boost the interaction, there are still two gaps between the visual and language domains: (1) the gap between the visual features and textual semantics, (2) the gap between the disordering of visual features and the ordering of texts.To bridge the gaps we propose a high-level semantic planning (HSP) mechanism that incorporates both a semantic reconstruction and an explicit order planning.We integrate the planning mechanism to the attention based caption model and propose the High-level Semantic PLanning based Attention Network (HS-PLAN).First, an attention based reconstruction module is designed to reconstruct the visual features with high-level semantic information.Then we apply a pointer network to serialize the features and obtain the explicit order plan to guide the generation.Experiments conducted on MS COCO show that our model outperforms previous methods and achieves the state-of-the-art performance of 133.4% CIDEr-D score.

Clothes category classification and attribute recognition have achieved distinguished success with the development of deep learning. People have found that landmark detection plays a positive role in these tasks. However, little research is committed to analyzing these tasks from the perspective of clothing attributes. In our work, we explore the usefulness of landmarks and find that landmarks can assist in extracting shape features; and using landmarks for joint learning can increase classification and recognition accuracy effectively. We also find that texture features have an impelling effect on these tasks and that the pre-trained ImageNet model has good performance in extracting texture features. To this end, we propose to use two streams to enhance the extraction of shape and texture, respectively. In particular, this paper proposes a simple implementation, Texture and Shape biased Fashion Networks (TS-FashionNet). Comprehensive and rich experiments demonstrate our discoveries and the effectiveness of our model. We improve the top-3 classification accuracy by 0.83% and improve the top-3 attribute recognition recall rate by 1.39% compared to the state-of-the-art models.

Weakly supervised localization has attracted increasing attention since only image-wise labels are needed. One mainstream approach, CAM based top-down localization method, suffers from poor resolution and localizing only the most discriminative regions. Another kind, model agnostic perturbation based method, suffers from multiple iterations for each sample. In this paper, we introduce PECA-Net: Preserve and Erase Based Class Attention Networks, which adopts preserve and erase perturbed U-net as the basis, with class activation mechanism as attention to enhance localization capability. Class attention module strengthens informative features and achieves a basic localization. Preserve and erase perturbed U-net replaces the random and iterative extrinsic perturbation with meaningful erasing. In addition, this structure refines the preliminary localization. Since the target object is hit twice, therefore, entitled as double shot. Experiments validate that localization error of both CUB-200 and ILSVRC ImageNet dataset is the new state-of-the-art.

Recently, Recurrent Neural Network (RNN) based methods and Self-Attention (SA) based methods have achieved promising performance in Video Question Answering (VideoQA). Despite the success of these works, RNN-based methods tend to forget the global semantic contents due to the inherent drawbacks of the recurrent units themselves, while SA-based methods cannot precisely capture the dependencies of the local neighborhood, leading to insufficient modeling for temporal order. To tackle these problems, we propose a novel VideoQA framework which progressively refines the representations of videos and questions from fine to coarse grain in a sequence-sensitive manner. Specifically, our model improves the feature representations via the following two steps: (1) introducing two fine-grained feature-augmented memories to strengthen the information augmentation of video and text which can improve memory capacity by memorizing more relevant and targeted information. (2) appending the self-attention and co-attention module to the memory output thus the module is able to capture global interaction between high-level semantic informations. Experimental results show that our approach achieves state-of-the-art performance on VideoQA benchmark datasets.

Stochastic video generation methods predict diverse videos based on observed frames, where the main challenge lies in modeling the complex future uncertainty and generating realistic frames. Numerous of Recurrent-VAE-based methods have achieved state-of-the-art results. However, on the one hand, the independence assumption of the variables of approximate posterior limits the inference performance. On the other hand, although these methods adopt skip connections between encoder and decoder to utilize multi-level features, they still produce blurry generation due to the spatial misalignment between encoder and decoder features at different time steps. In this paper, we propose a hierarchical recurrent VAE with a feature aligner, which can not only relax the independence assumption in typical VAE but also use a feature aligner to enable the decoder to obtain the aligned spatial information from the last observed frames. The proposed model is named Hierarchical Stochastic Video Generation network with Aligned Features, referred to as HAF-SVG. Experiments on Moving-MNIST, BAIR, and KTH datasets demonstrate that hierarchical structure is helpful for modeling more accurate future uncertainty, and the feature aligner is beneficial to generate realistic frames. Besides, the HAF-SVG exceeds SVG on both prediction accuracy and the quality of generated frames.

In transfer learning, what and where to transfer has been widely studied. Nevertheless, the learned transfer strategies are at high risk of over-fitting, especially when only a few annotated instances are available in the target domain. In this paper, we introduce the concept of transfer ensemble learning, a new direction to tackle the over-fitting of transfer strategies. Intuitively, models with different transfer strategies offer various perspectives on what and where to transfer. Therefore a core problem is to search these diversely transferred models for ensemble so as to achieve better generalization. Towards this end, we propose the Transferability Slider (TranSlider) for transfer ensemble learning. By decreasing the transferability, we obtain a spectrum of base models ranging from pure exploitation of the source model to unconstrained exploration for the target domain. Furthermore, the manner of decreasing transferability with parameter sharing guarantees fast optimization at no additional training cost. Finally, we conduct extensive experiments with various analyses, which demonstrate that TranSlider achieves the state-of-the-art on comprehensive benchmark datasets.

Scene graph generation aims to produce structured representations for images, which requires to understand the relations between objects. Due to the continuous nature of deep neural networks, the prediction of scene graphs is divided into object detection and relation classification. However, the independent relation classes cannot separate the visual features well. Although some methods organize the visual features into graph structures and use message passing to learn contextual information, they still suffer from drastic intra-class variations and unbalanced data distributions. One important factor is that they learn an unstructured output space that ignores the inherent structures of scene graphs. Accordingly, in this paper, we propose a Higher Order Structure Embedded Network (HOSE-Net) to mitigate this issue. First, we propose a novel structure-aware embedding-to-classifier(SEC) module to incorporate both local and global structural information of relationships into the output space. Specifically, a set of context embeddings are learned via local graph based message passing and then mapped to a global structure based classification space. Second, since learning too many context-specific classification subspaces can suffer from data sparsity issues, we propose a hierarchical semantic aggregation(HSA) module to reduces the number of subspaces by introducing higher order structural information. HSA is also a fast and flexible tool to automatically search a semantic object hierarchy based on relational knowledge graphs. Extensive experiments show that the proposed HOSE-Net achieves the state-of-the-art performance on two popular benchmarks of Visual Genome and VRD.

Learning related tasks in various domains and transferring exploited knowledge to new situations is a significant challenge in Reinforcement Learning (RL). However, most RL algorithms are data inefficient and fail to generalize in complex environments, limiting their adaptability and applicability in multi-task scenarios. In this paper, we propose SelfSupervised Mixture-of-Experts (SUM), an effective algorithm driven by predictive uncertainty estimation for multitask RL. SUM utilizes a multi-head agent with shared parameters as experts to learn a series of related tasks simultaneously by Deep Deterministic Policy Gradient (DDPG). Each expert is extended by predictive uncertainty estimation on known and unknown states to enhance the Q-value evaluation capacity against overfitting and the overall generalization ability. These enable the agent to capture and diffuse the common knowledge across different tasks improving sample efficiency in each task and the effectiveness of expert scheduling across multiple tasks. Instead of task-specific design as common MoEs, a self-supervised gating network is adopted to determine a potential expert to handle each interaction from unseen environments and calibrated completely by the uncertainty feedback from the experts without explicit supervision. To alleviate the imbalanced expert utilization as the crux of MoE, optimization is accomplished via decayedmasked experience replay, which encourages both diversification and specialization of experts during different periods. We demonstrate that our approach learns faster and achieves better performance by efficient transfer and robust generalization, outperforming several related methods on extended OpenAI Gym’s MuJoCo multi-task environments.

The bird community in the mangrove areas is an important component of the mangrove wetlands ecosystem and an indicator species for the assessment of the environmental health status of mangrove wetlands. The classification of bird species by the sound of bird in the mangrove areas has the advantages of less interference to the environment and wide monitoring range. In this paper, we propose a novel method that combines the feature recalibration mechanism with depthwise separable convolution for the mangrove bird sound classification. In the proposed method, we introduce Xception network in which depthwise separable convolution with lower parameter number and computational cost than traditional convolution can be stacked in a residual manner, as the baseline network. And we fuse the feature recalibration mechanism into the depthwise separable convolution for actively learning the weights of the feature channels in the network layer, so that we can enhance the important features in bird sound signals to improve the performance of the classification. In the proposed method, firstly we extract three-channel log-mel features of the bird sound signals and we introduce the mixup method to augment the extracted features. Secondly, we construct the recalibrated feature maps including the different scales of information to get the classification results. To verify the effectiveness of the proposed method, we build a dataset with 9282 samples including 25 kinds of the mangrove birds such as Egretta alba, Parus major, Charadrius dubius, etc. habiting in the mangroves of Fangcheng Port of China, and execute the experiments on the built dataset. Furthermore, we also validate the adaptability of our proposed method on the dataset of TAU Urban Acoustic Scenes 2019, and achieve a better result.

Many types of 3D acquisition sensors have emerged in recent years and point cloud has been widely used in many areas. Accurate and fast registration of cross-source 3D point clouds from different sensors is an emerged research problem in computer vision. This problem is extremely challenging because cross-source point clouds contain a mixture of various variances, such as density, partial overlap, large noise and outliers, viewpoint changing. In this paper, an algorithm is proposed to align cross-source point clouds with both high accuracy and high efficiency. There are two main contributions: firstly, two components, the weak region affinity and pixel-wise refinement, are proposed to maintain the global and local information of 3D point clouds. Then, these two components are integrated into an iterative tensor-based registration algorithm to solve the cross-source point cloud registration problem. We conduct experiments on a synthetic cross-source benchmark dataset and real cross-source datasets. Comparison with six state-of-the-art methods, the proposed method obtains both higher efficiency and accuracy.

Frame-to-frame uncertainty is a major challenge in video prediction. The use of the deterministic models always leads to averaging of future states. Some methods draw samples from a prior at each time step to deal with the uncertainty of the future states, such as the SVG model [1]. However, these models always use only one set of latent variables to represent the whole stochastic part in a video clip whereas sequential data often involves multiple independent factors. In this paper, we exploit the complex representation of information in video sequences by formulating it explicitly with a disentangled-representation stochastic video generation (DR-SVG) model that imposes sequence-dependent prior and sequence-independent prior to different sets of latent variables. Through a variational lower-bound and adversarial objective functions in latent space, our model can produce crisper frames with clear content and pose which indicate the sequence-dependent and sequence-independent component respectively.

Automatically generating natural language descriptions of images is a fundamental problem in artificial intelligence that connects computer vision and natural language processing. In recent years tremendous success has been shown in image captioning under the encoder-decoder framework, in which decoders are often chain-structured with Recurrent Neural Networks(RNNs), treating sentences as sequences. However, natural sentences are not inherently linear structures, but hierarchical structures. In this paper, we for the first time proposed a model with tree-structured decoder for image captioning(Image-to-Tree), which does not directly generate sentences but instead explicitly generates their dependency trees in a top-down manner. Inspired by the success of attention mechanism in image captioning, we also proposed a corresponding attention-based model for Image-to-Tree. Experiments on MSCOCO dataset demonstrate that our model can achieve comparable results to chain-structured models of different language metrics.

Preserving projection-based methods are good for finding the manifold structure embedded in data. As they use the Euclidean distance as a metric, which is sensitive to noise and outliers in data, nuclear norm-based 2D locality preserving projection (NN-2DLPP) is thus proposed to improve the robustness of 2DLPP. However, NN-2DLPP does not consider the discriminant ability of data. In order to improve the discriminant ability of preserving projection methods, in this paper, we use preserving projection learning with structurally incoherence of data and propose structurally incoherent low-rank 2DLPP (SILR-2DLPP) for image classification. This approach provides a discriminative representation of preserving projection learning by recovering the distinct different classes of the data. SILR-2DLPP searches the optimal subspace and low-rank representation simultaneously. We further extend SILR-2DLPP to a kernel case and propose kernel SILR-2DLPP (KSILR-2DLPP) to obtain a nonlinear representation. The theoretical analysis including the convergence and computational complexity of SILR-2DLPP are presented. To verify the performance of SILR-2DLPP and KSILR-2DLPP, six well-known image databases were used in the experiments. The experimental results show that the proposed methods are superior to the previous preserving projection methods for image classification.

2-D linear discriminant analysis (2DLDA) has been widely used in pattern recognition and image classification. 2DLDA selects discriminative features from the up and left corner of images. However, 2DLDA uses the Frobenius norm (F-norm), which is sensitive to noise or outliers in data, as a metric. In this paper, we propose a novel framework, called horizontal and vertical nuclear norm-based 2DLDA (HVNN-2DLDA) for image representation. In the proposed framework, HVNN-2DLDA methods (i.e., HNN-2DLDA and VNN-2DLDA) are proposed, and both use the nuclear norm as a criterion. The nuclear norm can provide more structure and global information for the reconstruction of noisy images. HNN-2DLDA and VNN-2DLDA represent images in the row and column directions, respectively. In addition, by combining the row and column directions, we propose a bilateral nuclear norm-based 2DLDA method called BNN-2DLDA. The advantage of BNN-2DLDA over HNN-2DLDA and VNN-2DLDA is that an image sample can be represented by both the row and the column directions instead of only the row or column direction. HVNN-2DLDA learns a set of local optimal projection vectors by maximizing the ratio of the nuclear norm of the between-class scatter matrix and the nuclear norm of the within-class scatter matrix. To verify the robustness and recognition performance in image classification of HVNN-2DLDA, six public image databases are used for experiments. The experimental results demonstrate the effectiveness and the feasibility of the proposed framework.

2-D neighborhood preserving projection (2DNPP) uses 2-D images as feature input instead of 1-D vectors used by neighborhood preserving projection (NPP). 2DNPP requires less computation time than NPP. However, both NPP and 2DNPP use the L2norm as a metric, which is sensitive to noise in data. In this paper, we proposed a novel NPP method called low-rank 2DNPP (LR-2DNPP). This method divided the input data into a component part that encoded low-rank features, and an error part that ensured the noise was sparse. Then, a nearest neighbor graph was learned from the clean data using the same procedure as 2DNPP. To ensure that the features learned by LR-2DNPP were optimal for classification, we combined the structurally incoherent learning and low-rank learning with NPP to form a unified model called discriminative LR-2DNPP (DLR2DNPP). By encoding the structural incoherence of the learned clean data, DLR-2DNPP could enhance the discriminative ability for feature extraction. Theoretical analyses on the convergence and computational complexity of LR-2DNPP and DLR-2DNPP were presented in details. We used seven public image databases to verify the performance of the proposed methods. The experimental results showed the effectiveness of our methods for robust image representation.

Combining Convolutional Neural Networks (CNNs) with Conditional Random Fields (CRFs) achieves great success among recent object segmentation methods. There are two advantages by such usage. First, CNNs can extract low-level features, which are very similar to the extracted features in primates' primary visual cortex (V1). Second, CRFs can set up the relationship between input features and output labels in a direct way. In this paper, we extend the first advantage by using CNNs for low-level feature extraction and a Structured Random Forest (SRF)-based border ownership detector for high-level feature extraction, which are similar to the outputs of primates secondary visual cortex (V2). Compared to the CRF model, an improved Conditional Boltzmann Machine (CBM), which has a multi-channel visible layer, is proposed to model the relationship between predicted labels, local and global contexts of objects with multi-scale and multilevel features. Besides, our proposed CBM model is extended for object parsing by using multivisible branches instead of a single visible layer of CBM, which cannot only segment the whole body but also the parts of the body under. These visible branches use each branch for the segmentation of the whole body or one of the body parts. All branches share the same hidden layers of CBM and train the branches under an iterative way. By exploiting object parsing, the whole body segmentation performance of object is improved. To refine the segmentation output, two kinds of optimization algorithms are proposed. The superpixel-based algorithm can re-label the overlapped regions of multiple kinds of objects. The other curve correction algorithm corrects the edges of segmented object parts by using smooth edges under a curve similarity criterion. Experiments demonstrate that our models yield competitive results for object segmentation on the PASCAL VOC 2012 dataset and for object parsing on the PennFudan Pedestrian Parsing dataset, Pedestrian Parsing Surveillance Scenes dataset, Horse-Cow parsing dataset, and PASCAL Quadrupeds dataset.

Existing visual tracking methods face many challenges: 1) the changed size and number of targets over time, occlusion in discrete frames, and mis-identification for crossing targets. Long short-term memory (LSTM) has the advantage of modeling long-term tasks and is suitable for tracking. We propose a novel online attentional recurrent neural network (ARNN) model for visual tracking, whose core component is a two-layer bidirectional LSTM along the x-and y-axes. Several bidirectional LSTMs can be cascaded or parallelly connected together to exploit multiscale target features and can give more precise tracked object locations. Each bidirectional LSTM utilizes the convolutional features of a convolutional neural network inside two bounding boxes from two frames to check whether the target in the current frame is the one in previous frames. An attention mechanism is also adopted to enhance the proposed model to better express the patch-level features of the tracking targets. Interattention and intra-attention models are proposed to imitate the temporal and spatial tracking mechanism of primate visual cortex. Interattention learns to overcome the occlusion problem, and intra-attention is able to mark important regions to better trace the target. The bidirectional LSTM and the attention mechanism are jointly trained. The combination of them further improves the accuracy of target tracking in videos. The outstanding performances in the experiments demonstrate the effectiveness of our proposed online method ARNN and yield competitive results compared with the state-of-the-art tracking methods.

The discriminative correlation filter (DCF) method is one of the main-stream methods in visual tracking with good performance and fast computational speed. However, most existing DCF trackers do not pay much attention to track a target in the low-resolution (LR) image. In order to address this problem, we propose a novel super-resolution tracker (SRT) based on DCF. SRT adopts a super-resolution reconstruction algorithm VDSR to enhance the resolution of the target. We compared SRT with 7 well-known DCF based trackers on 8 LR sequences selected from OTB2013. The results show that SRT outperforms state-of-the-art trackers in both quantitative and qualitative evaluations.

In this paper, we aim to tackle the task of captioning each event in one video (dense captioning in videos) and propose a novel pipeline. The task is challenging because of the uncertainty of event measurement as well as generating context-aware sentences. Conventional video captioning methods have flaws in encoding events with context, for a simple example: models would not correctly use words such as “another” and “continue” when describing multiple events. We directly deal with this issue by coming up with an encoder working along the time axis, which encodes videos and outputs features from different levels of hierarchical LSTMs. Our hierarchical LSTMs use different layers to retrieve intra-event (regional) and inter-event (global) descriptor. Moreover, we modify the language model in an attention fashion. Unlike previous attention modules, our attention module deals with regional and global information at different phases and integrates output into language LSTM. This results in better consistency with ground truth natural language. Besides, our attention model makes the whole pipeline more robust to inaccurate event proposal. We evaluate our method on the dense captioning dataset in common metrics, reporting as at most 30% boost on a single metric.

Recent work has shown that GANs can be stacked to generate plausible images, with each GAN generating lower-level representations conditioned on higher-level ones. However, the gap between generated representations and the real ones grows rapidly as the number of stacked GANs increases. In this paper, we aim to alleviate this problem and propose a novel generative model named DSGAN, which densely connects all stacks of GANs in order to provide enhanced conditional information for individual GANs. Besides, we introduce a weighted feature loss during training, which encourages representations reuse and provides multi-level supervision for each generator. Compared to existing architectures, our model maximizes internal information flow and prominently improves the discriminability of generated representations. We evaluate DSGAN on three public datasets: MNIST, SVHN and CIFAR-10. Experiments show that DSGAN outperforms current stacked GANs in generating realistic images, which has achieved obvious improvements on quantitative measures like Inception Score.

Thermal images are essential to deal with situations in dark environments, as they capture the objects' temperature. While the objects can still be seen in thermal images, the texture is extremely blur or even not observable at all. We propose to extract different features from images that capture various characteristics of the images. As one feature emphasizes one distinguishing aspect differing from the others, we can grasp multiple pieces of evidence from the images and take advantage of each to improve the thermal image classification accuracy. In particular, in additional to corner features usually used in color images, we also extract features from the edges and the shapes of the objects that emphasize the integral image appearance, as well as the temperature characteristics obtained from the image intensity. In this way, even if one feature is not evident in an image, the others can still play a critical role towards the correct classification result. By optimizing the objective function, we maximize the fusion performance of multiple features. By doing so, we can to the largest extent make use of the information exhibited in the thermal images to classify the query image into the correct group. Experiments demonstrate promising thermal image classification result.

Recursive neural network (RvNN) has been proved to be an effective and promising tool to learn sentence representations by explicitly exploiting the sentence structure. However, most existing work can only exploit simple tree structure, e.g., binary trees, or ignore the order of nodes, which yields suboptimal performance. In this paper, we proposed a novel neural network, namely TreeNet, to capture sentences structurally over the raw unconstrained constituency trees, where the number of child nodes can be arbitrary. In TreeNet, each node is learning from its left sibling and right child in a bottom-up left-to-right order, thus enabling the net to learn over any tree. Furthermore, multiple soft gates and a memory cell are employed in implementing the TreeNet to determine to what extent it should learn, remember and output, which proves to be a simple and efficient mechanism for semantic synthesis. Moreover, TreeNet significantly suppresses convolutional neural networks (CNN) and Long Short-Term Memory (LSTM) with fewer parameters. It improves the classification accuracy by 2%-5% with 42% of the best CNN’s parameters or 94% of standard LSTM’s. Extensive experiments demonstrate TreeNet achieves the state-of-the-art performance on all four typical text classification tasks.

Deep Deterministic Policy Gradient (DDPG) algorithm has been successful for state-of-the-art performance in high-dimensional continuous control tasks. However, due to the complexity and randomness of the environment, DDPG tends to suffer from inefficient exploration and unstable training. In this work, we propose Self-Adaptive Double Bootstrapped DDPG (SOUP), an algorithm that extends DDPG to bootstrapped actor-critic architecture. SOUP improves the efficiency of exploration by multiple actor heads capturing more potential actions and multiple critic heads evaluating more reasonable Q-values collaboratively. The crux of double bootstrapped architecture is to tackle the fluctuations in performance, caused by multiple heads of spotty capacity varying throughout training. To alleviate the instability, a self-adaptive confidence mechanism is introduced to dynamically adjust the weights of bootstrapped heads and enhance the ensemble performance effectively and efficiently. We demonstrate that SOUP achieves faster learning by at least 45% while improving cumulative reward and stability substantially in comparison to vanilla DDPG on OpenAI Gym's MuJoCo environments.

We propose an efficient method to deal with the matching and registration problem found in cross-source point clouds captured by different types of sensors. This task is especially challenging due to the presence of density variation, scale difference, a large proportion of noise and outliers, missing data, and viewpoint variation. The proposed method has two stages: in the coarse matching stage, we use the ensemble of shape functions descriptor to select potential K regions from the candidate point clouds for the target. In the fine stage, we propose a scale embedded generative Gaussian mixture models registration method to refine the results from the coarse matching stage. Following the fine stage, both the best region and accurate camera pose relationships between the candidates and target are found. We conduct experiments in which we apply the method to two applications: one is 3D object detection and localization in street-view outdoor (LiDAR/VSFM) cross-source point clouds and the other is 3D scene matching and registration in indoor (KinectFusion/VSFM) cross-source point clouds. The experiment results show that the proposed method performs well when compared with the existing methods. It also shows that the proposed method is robust under various sensing techniques, such as LiDAR, Kinect, and RGB camera.

Nonnegative matrix factorization (NMF), known as a famous matrix factorization technique, has been widely used in pattern recognition and computer vision. NMF represents the input data matrix as a product of two nonnegative factors. As NMF is based on the Euclidean distance, which is sensitive to noise or errors in the data, some robust NMF methods are proposed. Mainly focusing on parts-based representation, these robust NMF methods often neglect global representation of data. In fact, the global geometry information of data is more robust than the local information about the noisy data in terms of image classification. In order to effectively improve the robustness of NMF and learn part-based and global representation of the data, a novel method low-rank nonnegative factorization (LRNF) is proposed in this paper. First, we assume that the data are grossly corrupted, and the$L_{1} $norm is used as a sparse constraint on the assumed noise matrix. Then, LRNF learns a low-rank matrix with the global representation ability. Finally, we make a nonnegative factorization of the learned low-rank matrix. We can obtain a base matrix, which preserves locality and globality properties of the data in the meantime. Extensive experiments have been conducted on nine real-world image databases to verify the performance of the proposed LRNF method by comparing with the state-of-the-art algorithms on robust dimensionality reduction.

As a popular dimensionality reduction method, nonnegative matrix factorization (NMF) has been widely used in image classification. However, the NMF does not consider discriminant information from the data themselves. In addition, most NMF-based methods use the Euclidean distance as a metric, which is sensitive to noise or outliers in data. To solve these problems, in this paper, we introduce structural incoherence and low-rank to NMF and propose a novel nonnegative factorization method, called structurally incoherent low-rank NMF (SILR-NMF), in which we jointly consider structural incoherence and low-rank properties of data for image classification. For the corrupted data, we use the norm as a constraint to ensure the noise is sparse. SILR-NMF learns a clean data matrix from the noisy data by low-rank learning. As a result, the SILR-NMF can capture the global structure information of the data, which is more robust than local information to noise. By introducing the structural incoherence of the learned clean data, SILR-NMF ensures the clean data points from different classes are as independent as possible. To verify the performance of the proposed method, extensive experiments are conducted on six image databases. The experimental results demonstrate that our proposed method has substantial gain over existing NMF approaches.

The aim of scene semantic segmentation is to label each pixel with a class which it belongs to in high level cognition. State-of-art works mainly adapt convolutional neural networks originally designed for image classification to make dense prediction. However the inner structure of scene itself and its stuff is more flexible and variable, which is distinct from the objects in image classification task. Therefore we propose a gate function based structure-aware convolution for deep neural networks with the ability of modeling inner variance in scene. The gate function is a RNN-based learnable function or a handcrafted one, which is applied to distinguish efficient activations from convolution area. It is proved that dilated convolution is a subclass of gate function. As shown in our experiments on scene datasets, the proposed convolution method efficiently improves the accuracy of current semantic segmentation systems by partly replacing original networks' convolution layers with ours.

We propose a novel online Attentional Recurrent Neural Network (ARNN) model for visual tracking, which exploits the feature maps of Convolutional Neural Network (CNN) inside a bounding box to identify whether this target is the one appeared in previous frames. Attention mechanism is adopted for both different parts of targets and different scales of object features. The former attention model is able to select important regions to better trace the target while the latter one learns to weight the multiple scale features for accurate object location. We jointly train the recurrent network with the region based and scale based attention mechanism. The outstanding performances in the experiments validate the effectiveness of our proposed ARNN and show that ARNN outperforms the state-of-the-art tracking methods.

Nonnegative matrix factorization (NMF), which aims at obtaining the nonnegative low-dimensional representation of data, has received wide attention. To obtain more effective nonnegative discriminant bases from the original NMF, in this paper, a novel method called nonnegative discriminant matrix factorization (NDMF) is proposed for image classification. NDMF integrates the nonnegative constraint, orthogonality, and discriminant information in the objective function. NDMF considers the incoherent information of both factors in standard NMF and is proposed to enhance the discriminant ability of the learned base matrix. NDMF projects the low-dimensional representation of the subspace of the base matrix to regularize the NMF for discriminant subspace learning. Based on the Euclidean distance metric and the generalized Kullback-Leibler (KL) divergence, two kinds of iterative algorithms are presented to solve the optimization problem. The between- and within-class scatter matrices are divided into positive and negative parts for the update rules and the proofs of the convergence are also presented. Extensive experimental results demonstrate the effectiveness of the proposed method in comparison with the state-of-the-art discriminant NMF algorithms.

In this paper, a novel classifier, called superimposed sparse parameter (SSP) classifier is proposed for face recognition. SSP is motivated by two phase test sample sparse representation (TPTSSR) and linear regression classification (LRC), which can be treated as the extended of sparse representation classification (SRC). SRC uses all the train samples to produce the sparse representation vector for classification. The LRC, which can be interpreted as L2-norm sparse representation, uses the distances between the test sample and the class subspaces for classification. TPTSSR is also L2-norm sparse representation and uses two phase to compute the distance for classification. Instead of the distances, the SSP classifier employs the SSPs, which can be expressed as the sum of the linear regression parameters of each class in iterations, is used for face classification. Further, the fast SSP (FSSP) classifier is also suggested to reduce the computation cost. A mass of experiments on Georgia Tech face database, ORL face database, CVL face database, AR face database, and CASIA face database are used to evaluate the proposed algorithms. The experimental results demonstrate that the proposed methods achieve better recognition rate than the LRC, SRC, collaborative representation-based classification, regularized robust coding, relaxed collaborative representation, support vector machine, and TPTSSR for face recognition under various conditions.

We propose a systematic approach for registering cross-source point clouds that come from different kinds of sensors. This task is especially challenging due to the presence of significant missing data, large variations in point density, scale difference, large proportion of noise, and outliers. The robustness of the method is attributed to the extraction of macro and micro structures. Macro structure is the overall structure that maintains similar geometric layout in cross-source point clouds. Micro structure is the element (e.g., local segment) being used to build the macro structure. We use graph to organize these structures and convert the registration into graph matching. With a novel proposed descriptor, we conduct the graph matching in a discriminative feature space. The graph matching problem is solved by an improved graph matching solution, which considers global geometrical constraints. Robust cross source registration results are obtained by incorporating graph matching outcome with RANSAC and ICP refinements. Compared with eight state-of-the-art registration algorithms, the proposed method invariably outperforms on Pisa Cathedral and other challenging cases. In order to compare quantitatively, we propose two challenging cross-source data sets and conduct comparative experiments on more than 27 cases, and the results show we obtain much better performance than other methods. The proposed method also shows high accuracy in same-source data sets.

Two-dimensional locality preserving projections (2DLPP) that use 2D image representation in preserving projection learning can preserve the intrinsic manifold structure and local information of data. However, 2DLPP is based on the Euclidean distance, which is sensitive to noise and outliers in data. In this paper, we propose a novel locality preserving projection method called nuclear norm-based two-dimensional locality preserving projections (NN-2DLPP). First, NN-2DLPP recovers the noisy data matrix through low-rank learning. Second, noise in data is removed and the learned clean data points are projected on a new subspace. Without the disturbance of noise, data points belonging to the same class are kept as close to each other as possible in the new projective subspace. Experimental results on six public image databases with face recognition, object classification, and handwritten digit recognition tasks demonstrated the effectiveness of the proposed method.

In this paper, a new classifier, called multiple linear regression coefficients (MLRC), is proposed for image recognition. Linear regression classification (LRC) uses the linear combination of the class-model for classification. Sparse representation based classification (SRC) utilizes the globalmodel for classification. Based on the global-concept of SRC, mean representation classification (MRC) is proposed, which uses the mean vector of each class to constitute the global-model for classification. Motivated by LRC, MRC and SRC, this paper focuses on finding the better global-model for classification. MLRC firstly constitutes regression coefficient matrix and multiple sub-global-models. Afterwards, MLRC computes the weighted value with the coefficient matrix and solve the least square error with multiple sub-global-models for classification. The PolyU Finger-Knuckle-Print (FKP) and Multispectral palm-print (MSP) databases are used to assess the proposed classifier. Experimental results demonstrate that the proposed approach achieves a better recognition rate than the LRC, MRC, SRC and some state-of-the-art methods.

Computing matching cost by Convolutional neural networks(CNNs) work well in fetching accurate dense disparity maps. But these methods still have problems: (1) they always employ equal weights for left and right images in convolutional layers, losing relational information of patches; (2) they don't solve the balance between patches' size and processing efficiency, the larger size the more information but slower. The proposed multi-scale CNN structured method fetches contexts by employing down-sampled images, this increases the matching accuracy without enlarging the input patch. A multi-scale cross-based aggregation algorithm is proposed to further refine the performances. Proposed method achieves challenging performance against state-of-the-art methods with an error rate of 4.77% in KITTI non-occluded section.

Existing deep neural networks, such as Convolutional Neural Networks (CNNs) and Recurrent Neural Networks (RNNs), typically treat volumetric video data as several single images and deal with one frame at one time, thus the relevance to frames can hardly be fully exploited. Besides, depth context plays the unique role in motion scenes for primates, but is seldom used in no depth label situations. In this paper, we use a more suitable architecture Multi-Scale Pyramidal Multi-Dimensional Long Short Term Memory (MSPMD-LSTM) to reveal the strong relevance within video frames. Furthermore, depth context is extracted and refined to enhance the performance of the model. Experiments demonstrate that our models yield competitive results on Youtube-Objects dataset and Segtrack v2 dataset.

Existing joint models of deep Convolutional Neural Networks (CNNs) and Conditional Random Fields (CRFs) face two problems for object segmentation: 1) CNNs can hardly extract high level features; 2) fully connected layers of CNNs are lack of capability of dealing with structured multi-level features. To address these problems, we utilize a Structured Random Forests based border ownership detection method to extract high level border features, which simulates the function of humans secondary visual cortex (V2). Moreover, an improved Conditional Boltzmann Machines (CBMs) are proposed to model predicted labels, local and global contexts of objects with multi-scale and multilevel features. Meanwhile, the proposed model inherits the merits of CNN, i.e., the good simulation of low level feature extraction ability in primary visual cortex (V1). Experiments demonstrate that our models yield competitive results on PASCAL VOC 2012 dataset.

We propose an expectation-maximization-like(EM-like) method to train Boltzmann machine with unconstrained connectivity. It adopts Monte Carlo approximation in the E-step, and replaces the intractable likelihood objective with efficiently computed objectives or directly approximates the gradient of likelihood objective in the M-step. The EM-like method is a modification of alternating minimization. We prove that EM-like method will be the exactly same with contrastive divergence in restricted Boltzmann machine if the M-step of this method adopts special approximation. We also propose a new measure to assess the performance of Boltzmann machine as generative models of data, and its computational complexity is O(Rmn). Finally, we demonstrate the performance of EM-like method using numerical experiments.

As one of the most popular dimensionality reduction techniques, locality preserving projections (LPP) has been widely used in computer vision and pattern recognition. However, in practical applications, data is always corrupted by noises. For the corrupted data, samples from the same class may not be distributed in the nearest area, thus LPP may lose its effectiveness. In this paper, it is assumed that data is grossly corrupted and the noise matrix is sparse. Based on these assumptions, we propose a novel dimensionality reduction method, named low-rank preserving projections (LRPP) for image classification. LRPP learns a low-rank weight matrix by projecting the data on a low-dimensional subspace. We use the L21 norm as a sparse constraint on the noise matrix and the nuclear norm as a low-rank constraint on the weight matrix. LRPP keeps the global structure of the data during the dimensionality reduction procedure and the learned low rank weight matrix can reduce the disturbance of noises in the data. LRPP can learn a robust subspace from the corrupted data. To verify the performance of LRPP in image dimensionality reduction and classification, we compare LRPP with the state-of-the-art dimensionality reduction methods. The experimental results show the effectiveness and the feasibility of the proposed method with encouraging results.

Recently, the linear regression classification (LRC) has been proposed. LRC uses linear regression representation of each class for classification. Based on LRC, a few improved methods are proposed, such as kernel linear regression classification (KLRC), kernel ridge regression classification (KRRC) and nearest regularized subspace (NRS). Motivated by the LRC, KLRC, KRRC and NRS, center-based weighted kernel linear regression (CWKLR) is proposed for image classification in this paper. CWKLR firstly constitutes the center-based kernel matrix based on the KLRC and KRRC. Next, CWKLR utilizes the Tikhonov Matrix to solve the weighted projection coefficients for classification. Experiments on Coil100 object database, Eth80 object database and GT face database are used to evaluate the proposed algorithms. The experimental results demonstrate that the proposed methods achieve better recognition rate than some state-of-the-art methods.

In a network, the location of the node on which a service is computed is inextricably linked to the locations of the paths through which the service communicates. Hence, service location can have a profound effect on quality of service, especially for communication-centric applications such as real-time multimedia. In this paper, we propose an online algorithm that uses pricing to consider server load, route congestion, and propagation delay jointly when locating servers and routes for real-time multimedia services in a network with fixed computing and communication capacities. The algorithm is online in the sense that it is able to sequentially allocate resources for services with long and unknown duration as demands arrive, without the benefit of looking ahead to later demands. By formulating the problem as one of lowest cost subgraph packing, we prove that our algorithm is neverthelessC-competitive with the optimal algorithm that looks ahead, meaning that our performance is within a constant factorCof optimal, as measured by the total number of service demands satisfied, or total user utility. Using mixing services as an example, we show through experimental results that our algorithm can adapt to cross traffic and automatically route around congestion and failure of nodes and edges, can reduce latency by 40% or more, and can pack 20% more sessions or alternatively can double the number of sessions before significant call rejection, compared with conventional approaches.

Traditional multiparty audio or video conferencing uses a single node, sometimes called a multipoint control unit, or MCU, to mix audio data for the conference. We introduce a novel mixer, called a Virtual Mixer, which performs mixing in a distributed way over the network. The Virtual Mixer topology is optimized over Steiner trees using a metric of either average pairwise delay (APD) or maximum pairwise delay (MPD). Since the topology is adapted to the particular set of clients and servers available in the cloud, optimization speed is important. In order to solve this NP-hard Steiner tree optimization, we propose heuristic algorithms for finding the Pairwise-delay-optimal Tree (PT) for both APD and MPD, which are orders of magnitude faster than exhaustive search, yet find trees with delays that are minimal or within a few percent of minimal. We show through experiments both on a corporate intranet and on up to 12 PlanetLab nodes that Virtual Mixing can reduce both the APD and the MPD between clients by upwards of 50%, compared with the existing MCU-based and P2P-based mixing approaches.

The scalable video coding is an effective solution to fulfil the different requirements in modern video transmission system. Though the coding efficiency is high, the computational complexity is expensive. This paper presents a proposed algorithm to reduce the encoding time of enhancement layer through probability analysis. Firstly two classes are defined, and the Bayes Classifier decides whether the current macroblock belongs to Big Block class or Small Block class. Then an early termination strategy is checked for ignoring some inter modes. When inter-layer residual prediction is used, the optimal inter mode is set as the best inter mode of co-located macroblock in base layer. For the intra modes, the INTRA_16×16 is skipped based on statistical data. Experiment results show that the proposed algorithm reduces computational complexity with negligible video quality loss and bit rate increment when compared with reference software and other methods.

Trace and revoke systems are motivated by content protection on various platforms such as DVD players, satellite radio receivers and PCs, which provide both broadcast encryption and traitor tracing mechanisms. So far, the best existing system requires ciphertext size sub-linear in the number of users. In communications, as system scale increases, it brings up a high bandwidth requirement which cannot be satisfied by current network condition. In this paper, we present a new fully collusion resistant trace and revoke system which has only constant size ciphertexts. The system is CPA secure against static adversaries with public broadcast key and secret tracing key. It can be efficiently applied in DRM fields where network bandwidth is limited.

This paper presents an online handwritten character recognition system. The whole system includes three parts: acceleration signal detection, signal processing and recognition by Hidden Markov Model (HMM). In hardware aspect, a mini-board with a three-dimensional accelerometer and a microcontroller

The newly adopted MPEG-4 fine granularity scalability (FGS) video coding standard offers easy and flexible adaptation to varying network bandwidths and different application needs. Encryption for FGS should preserve such adaptation capabilities and enable intermediate stages to process encrypted data directly without decryption. In this paper, we propose two novel encryption algorithms for MPEG-4 FGS that meet these requirements. The first algorithm encrypts an FGS stream (containing both the base and the enhancement layers) into a single access layer and preserves the original fine granularity scalability and error resilience performance in an encrypted stream. The second algorithm encrypts an FGS stream into multiple quality layers divided according to either peak signal-to-noise ratio (PSNR) or bit rates, with lower quality layers being accessible and reusable by a higher quality layer of the same type, but not vice versa. Both PSNR and bit-rate layers are supported simultaneously so a layer of either type can be selected on the fly without decryption. The base layer for the second algorithm may be unencrypted to allow free view of the content at low-quality or content-based search of a video database without decryption. Both algorithms are fast, error-resilient, and have negligible compression overhead. The same approach can be applied to other scalable multimedia formats.

MPEG-4 has recently adopted the fine granularity scalability (FGS) video coding technology which enables easy and flexible adaptation to bandwidth fluctuations and device capabilities. Encryption for FGS should preserve such adaptation capabilities and allow intermediate stages in the delivery to process the media on the ciphertext directly. In this paper, we propose a novel scalable access control scheme with this property for the MPEG-4 FGS format. It offers free browsing of the low-quality base layer video but controls the access to the enhancement layer at different service levels based on either PSNR or bitrates. Both types of service levels are supported simultaneously without jeopardizing each other's security. The scheme is fast and degrades neither compression efficiency nor error resilience of the MPEG-4 FGS. The approach is also applicable to other scalable multimedia.