With the rapid advancement of remote sensing technology, high-resolution multi-modal imagery is now more widely accessible. Conventional object detection models are trained on a single dataset, often restricted to a specific imaging modality and annotation format. However, such an approach overlooks the valuable shared knowledge across multi-modalities and limits the model’s applicability in more versatile scenarios. This paper introduces a new task called Multi-Modal Datasets and Multi-Task Object Detection (M2Det) for remote sensing, designed to accurately detect horizontal or oriented objects from any sensor modality. This task poses challenges due to 1) the trade-offs involved in managing multi-modal modelling and 2) the complexities of multi-task optimization. To address these, we establish a benchmark dataset and propose a unified model, SM3Det (Single Model for Multi-Modal datasets and Multi-Task object Detection). SM3Det leverages a grid-level sparse MoE backbone to enable joint knowledge learning while preserving distinct feature representations for different modalities. Furthermore, we propose a novel consistency and synchronization optimization mechanism, allowing it to effectively handle varying levels of learning difficulty across modalities and tasks. Extensive experiments demonstrate SM3Det's effectiveness and generalizability, consistently outperforming the combination of specialized models on individual datasets.

In this paper, we show that current approaches using large square kernels or transformer-based global modeling aggregate contextual information uniformly across spatial dimensions, leading to feature dilution and localization errors for elongated targets. To mitigate this issue, we propose Strip R-CNN, the first work to systematically explore large strip convolutions for remote sensing object detection. Our key insight is that strip convolutions enable directional feature aggregation along the dominant spatial dimension of slender objects, reducing background interference while preserving essential geometric information. We design two core components: (i) StripNet, a backbone network employing sequential orthogonal large strip convolutions to capture anisotropic spatial patterns, and (ii) Strip Head, which enhances localization precision by incorporating strip convolutions into the detection head. Unlike previous large-kernel approaches that suffer from computational redundancy and isotropic limitations, our method achieves superior performance with remarkable efficiency. Extensive experiments on multiple benchmarks (DOTA, FAIR1M, HRSC2016, and DIOR) demonstrate significant improvements, with our 30M parameter model achieving 82.75% mAP on DOTA-v1.0, establishing a new state-of-the-art record while providing new insights into anisotropic feature learning for remote sensing applications.

A significant research effort is focused on exploiting the outstanding capacities of pretrained diffusion models for image editing. Approaches either fine tune the model, or invert the image in the latent space of the pretrained model. However, they suffer from two problems: (i) unsatisfactory results in selected regions and unexpected changes in non selected regions, and (ii) the need for careful text prompt editing: the prompt should include all visual objects in the input image. To address this, we propose two improvements: (i) only optimizing the input of the value linear network in the cross-attention layers is sufficiently powerful to reconstruct a real image, and (ii) attention regularization to preserve the object-like attention maps after reconstruction and editing, enabling accurate style editing without causing significant structural change. We further improve the editing technique used for the unconditional branch of classifier-free guidance as used by P2P. Extensive experimental prompt-editing results on a variety of images demonstrate qualitatively and quantitatively that our method has editing capabilities superior to those of existing and concurrent works. Our StyleDiffusion code is available at https://github.com/sen-mao/StyleDiffusion.

We explore the potential of pretrain-and-finetune manner on the RGB-D semantic segmentation to solve the common mismatch problem in this field. Specifically, we present DFormer++, a novel RGB-D pretrain-and-finetune framework to learn transferable representations for RGB-D semantic segmentation. This paper has two vital innovations. 1) Framework perspective: Different from the existing methods that finetune RGB pretrained backbone to the RGB-D scenes, we pretrain the backbone using image-depth pairs from ImageNet-1 K, and hence the model is endowed with the capacity to encode RGB-D representations; 2) Architecture perspective: Our model comprises a sequence of RGB-D attention blocks, which are tailored for encoding both RGB and depth information through a novel attention mechanism. Our DFormer++ avoids the mismatched encoding of the 3D geometry relationships in depth maps by RGB pretrained backbones, which widely lies in previous works but has not been resolved. Meanwhile, the tailored architecture greatly reduces redundant parameters for encoding RGB-D data and achieves efficient and accurate perception. Experimental results show that our DFormer++ achieves new cutting-edge performance on three popular RGB-D semantic segmentation benchmarks.

Previous works have shown that increasing the window size for Transformer-based image super-resolution models (e.g., SwinIR) can significantly improve the model performance. Still, the computation overhead is also considerable when the window size gradually increases. In this paper, we present SRFormer, a simple but novel method that can enjoy the benefit of large window self-attention but introduces even less computational burden. The core of our SRFormer is the permuted self-attention (PSA), which strikes an appropriate balance between the channel and spatial information for self-attention. Without any bells and whistles, we show that our SRFormer achieves a 33.86 dB PSNR score on the Urban100 dataset, which is 0.46 dB higher than that of SwinIR but uses fewer parameters and computations. In addition, we also attempt to scale up the model by further enlarging the window size and channel numbers to explore the potential of Transformer-based models. Experiments show that our scaled model, named SRFormerV2, can further improve the results and achieves state-of-the-art. We hope our simple and effective approach could be useful for future research in super-resolution model design.

Despite significant progress in multimodal large language models (MLLMs), their performance on complex, multi-page document comprehension remains inadequate, largely due to the lack of high-quality, document-level datasets. While current retrieval-augmented generation (RAG) methods offer partial solutions, they suffer from issues, such as fragmented retrieval contexts, multi-stage error accumulation, and extra time costs of retrieval. In this work, we present a high-quality document-level dataset, Doc-750K, designed to support in-depth understanding of multimodal documents. This dataset includes diverse document structures, extensive cross-page dependencies, and real question-answer pairs derived from the original documents. Building on the dataset, we develop a native multimodal model—Docopilot, which can accurately handle document-level dependencies without relying on RAG. Experiments demonstrate that Docopilot achieves superior coherence, accuracy, and efficiency in document understanding tasks and multi-turn interactions, setting a new baseline for document-level multimodal understanding. Data, code, and models are released at https://github.com/OpenGVLab/Docopilot.

Continual learning requires models to train continuously across consecutive tasks without forgetting. Most existing methods utilize linear classifiers, which struggle to maintain a stable classification space while learning new tasks. Inspired by the success of Kolmogorov-Arnold Networks (KAN) in preserving learning stability during simple continual regression tasks, we set out to explore their potential in more complex continual learning scenarios. In this paper, we introduce the Kolmogorov-Arnold Classifier (KAC), a novel classifier developed for continual learning based on the KAN structure. We delve into the impact of KAN’s spline functions and introduce Radial Basis Functions (RBF) for improved compatibility with continual learning. We replace linear classifiers with KAC in several recent approaches and conduct experiments across various continual learning benchmarks, all of which demonstrate performance improvements, highlighting the effectiveness and robustness of KAC in continual learning. The code is available at https://github.com/Ethanhuhuhu/KAC.

Recent studies have explored the combination of different LoRAs to jointly generate learned style and content. However, existing methods either fail to effectively preserve both the original subject and style simultaneously or require additional training. In this paper, we argue that the intrinsic properties of LoRA can effectively guide diffusion models in merging learned subject and style. Based on this insight, we propose K-LoRA, a simple yet effective training-free LoRA fusion approach. In each attention layer, K-LoRA compares the Top-K elements in each LoRA to be fused, determining which LoRA to select for optimal fusion. This selection mechanism ensures that the most representative features of both subject and style are retained during the fusion process, effectively balancing their contributions. Experiments demonstrate that K-LoRA can effectively integrates the subject and style information learned by the original LoRAs, outperforming state-of-the-art training-based approaches in both qualitative and quantitative results.

Recent advances in scene understanding benefit a lot from depth maps because of the 3D geometry information, especially in complex conditions (e.g., low light and overexposed). Existing approaches encode depth maps along with RGB images and perform feature fusion between them to enable more robust predictions. Taking into account that depth can be regarded as a geometry supplement for RGB images, a straightforward question arises: Do we really need to explicitly encode depth information with neural networks as done for RGB images? Based on this insight, in this paper, we investigate a new way to learn RGBD feature representations and present DFormerv2, a strong RGBD encoder that explicitly uses depth maps as geometry priors rather than encoding depth information with neural networks. Our goal is to extract the geometry clues from the depth and spatial distances among all the image patch tokens, which will then be used as geometry priors to allocate attention weights in self-attention. Extensive experiments demonstrate that DFormerv2 exhibits exceptional performance in various RGBD semantic segmentation benchmarks. Code is available at: https://github.com/VCIP-RGBD/DFormer.

We present TAR3D, a novel framework that consists of a 3D-aware Vector Quantized-Variational AutoEncoder (VQ-VAE) and a Generative Pre-trained Transformer (GPT) to generate high-quality 3D assets. The core insight of this work is to migrate the multimodal unification and promising learning capabilities of the next-token prediction paradigm to conditional 3D object generation. To achieve this, the 3D VQ-VAE first encodes a wide range of 3D shapes into a compact triplane latent space and utilizes a set of discrete representations from a trainable codebook to reconstruct fine-grained geometries under the supervision of query point occupancy. Then, the 3D GPT, equipped with a custom triplane position embedding called TriPE, predicts the codebook index sequence with prefilling prompt tokens in an autoregressive manner so that the composition of 3D geometries can be modeled part by part. Extensive experiments on ShapeNet and Objaverse demonstrate that TAR3D can achieve superior generation quality over existing methods in text-to-3D and image-to-3D tasks

Semantic segmentation is fundamental to vision systems requiring pixel-level scene understanding, yet deploying it on resource-constrained devices demands efficient architectures. Although existing methods achieve real-time inference through lightweight designs, we reveal their inherent limitation: misalignment between class representations and image features caused by a per-pixel classification paradigm. With experimental analysis, we find that this paradigm results in a highly challenging assumption for efficient scenarios: Image pixel features should not vary for the same category in different images. To address this dilemma, we propose a coupled dual-branch offset learning paradigm that explicitly learns feature and class offsets to dynamically refine both class representations and spatial image features. Based on the proposed paradigm, we construct an efficient semantic segmentation network, OffSeg. Notably, the offset learning paradigm can be adopted to existing methods with no additional architectural changes. Extensive experiments on four datasets, including ADE20K, Cityscapes, COCO-Stuff-164K, and Pascal Context, demonstrate consistent improvements with negligible parameters. For instance, on the ADE20K dataset, our proposed offset learning paradigm improves SegFormer-B0, SegNeXt-T, and Mask2Former-Tiny by 2.7%, 1.9%, and 2.6% mIoU, respectively, with only 0.1-0.2M additional parameters required.

Language-based object detection (LOD) aims to align visual objects with language expressions. A large amount of paired data is utilized to improve LOD model generalizations. During the training process, recent studies leverage vision-language models (VLMs) to automatically generate human-like expressions for visual objects, facilitating training data scaling up. In this process, we observe that VLM hallucinations bring inaccurate object descriptions (e.g., object name, color, and shape) to deteriorate VL alignment quality. To reduce VLM hallucinations, we propose an agentic workflow controlled by an LLM to re-align language to visual objects via adaptively adjusting image and text prompts. We name this workflow Real-LOD, which includes planning, tool use, and reflection steps. Given an image with detected objects and VLM raw language expressions, Real-LOD reasons its state automatically and arranges action based on our neural symbolic designs (i.e., planning). The action will adaptively adjust the image and text prompts and send them to VLMs for object re-description (i.e., tool use). Then, we use another LLM to analyze these refined expressions for feedback (i.e., reflection). These steps are conducted in a cyclic form to gradually improve language descriptions for re-aligning to visual objects. We construct a dataset that contains a tiny amount of 0.18M images with re-aligned language expression and train a prevalent LOD model to surpass existing LOD methods by around 50% on the standard benchmarks. Our Real-LOD workflow, with automatic VL refinement, reveals a potential to preserve data quality along with scaling up data quantity, which further improves LOD performance from a data-alignment perspective.

Diffusion models have exhibited extraordinary performance in dense prediction tasks. However, there are few works exploring the diffusion pipeline for multi-task dense predictions. In this paper, we unlock the potential of diffusion models in solving multi-task dense predictions and propose a novel diffusion-based method, called TaskDiffusion, which leverages the conditional diffusion process in the decoder. Instead of denoising the noisy labels for different tasks separately, we propose a novel joint denoising diffusion process to capture the task relations during denoising. To be specific, our method first encodes the task-specific labels into a task-integration feature space to unify the encoding strategy. This allows us to get rid of the cumbersome task-specific encoding process. In addition, we also propose a cross-task diffusion decoder conditioned on task-specific multi-level features, which can model the interactions among different tasks and levels explicitly while preserving efficiency. Experiments show that our TaskDiffusion outperforms previous state-of-the-art methods for all dense prediction tasks on the widely-used PASCAL-Context and NYUD-v2 datasets. Our code is available at https://github.com/YuqiYang213/TaskDiffusion.

This paper introduces TempSamp-R1, a new reinforcement fine-tuning framework designed to improve the effectiveness of adapting multimodal large language models (MLLMs) to video temporal grounding tasks. We reveal that existing reinforcement learning methods, such as Group Relative Policy Optimization (GRPO), rely on on-policy sampling for policy updates. However, in tasks with large temporal search spaces, this strategy becomes both inefficient and limited in performance, as it often fails to identify temporally accurate solutions. To address this limitation, TempSamp-R1 leverages ground-truth annotations as off-policy supervision to provide temporally precise guidance, effectively compensating for the sparsity and misalignment in on-policy solutions. To further stabilize training and reduce variance in reward-based updates, TempSamp-R1 provides a non-linear soft advantage computation method that dynamically reshapes the reward feedback via an asymmetric transformation. By employing a hybrid Chain-of-Thought (CoT) training paradigm, TempSamp-R1 optimizes a single unified model to support both CoT and non-CoT inference modes, enabling efficient handling of queries with varying reasoning complexity. Experimental results demonstrate that TempSamp-R1 outperforms GRPO-based baselines, establishing new state-of-the-art performance on benchmark datasets:Charades-STA (R1\@0.7: 52.9\%, +**2.7**\%), ActivityNet Captions (R1\@0.5: 56.0\%, +**5.3**\%), and QVHighlights (mAP: 30.0\%, +**3.0**\%). Moreover, TempSamp-R1 shows robust few-shot generalization capabilities under limited data. Code is available at https://github.com/HVision-NKU/TempSamp-R1.

Recent research on representation learning has proved the merits of multi-modal clues for robust semantic segmentation. Nevertheless, a flexible pretrain-and-finetune pipeline for multiple visual modalities remains unexplored. In this paper, we propose a novel multi-modal learning framework, termed OmniSegmentor. It has two key innovations: 1) Based on ImageNet, we assemble a large-scale dataset for multi-modal pretraining, called OmniSegmentor, which contains five popular visual modalities; 2) We provide an efficient pretraining manner to endow the model with the capacity to encode different modality information in the OmniSegmentor. For the first time, we introduce a universal multi-modal pretraining framework that consistently amplifies the model's perceptual capabilities across various scenarios, regardless of the arbitrary combination of the involved modalities. Remarkably, our OmniSegmentor achieves new state-of-the-art records on a wide range of multi-modal semantic segmentation datasets, including NYU Depthv2, EventScape, MFNet, DeLiVER, SUNRGBD, and KITTI-360. Data, model checkpoints, and source code will be made publicly available: https://github.com/VCIP-RGBD/DFormer.

Graphical User Interface (GUI) agents have made substantial strides in understanding and executing user instructions across diverse platforms. Yet, grounding these instructions to precise interface elements remains challenging—especially in complex, high-resolution, professional environments. Traditional supervised fine-tuning (SFT) methods often require large volumes of diverse data and exhibit weak generalization. To overcome these limitations, we introduce a reinforcement learning (RL)-based framework that incorporates three core strategies: (1) seed data curation to ensure high-quality training samples, (2) a dense policy gradient that provides continuous feedback based on prediction accuracy, and (3) a self-evolutionary reinforcement finetuning mechanism that iteratively refines the model using attention maps. With only 3k training samples, our 7B-parameter model achieves state-of-the-art results among similarly sized models on three grounding benchmarks. Notably, it attains 47.3\% accuracy on the ScreenSpot-Pro dataset—outperforming much larger models, such as UI-TARS-72B, by a margin of 24.2\%. These findings underscore the effectiveness of RL-based approaches in enhancing GUI agent performance, particularly in high-resolution, complex environments.

We aim at providing the object detection community with an efficient and performant object detector, termed YOLO-MS. The core design is based on a series of investigations on how multi-branch features of the basic block and convolutions with different kernel sizes affect the detection performance of objects at different scales. The outcome is a new strategy that can significantly enhance multi-scale feature representations of real-time object detectors. To verify the effectiveness of our work, we train our YOLO-MS on the MS COCO dataset from scratch without relying on any other large-scale datasets, like ImageNet or pre-trained weights. Without bells and whistles, our YOLO-MS outperforms the recent state-of-the-art real-time object detectors, including YOLO-v7, RTMDet, and YOLO-v8. Taking the XS version of YOLO-MS as an example, it can achieve an AP score of 42+% on MS COCO, which is about 2% higher than RTMDet with the same model size. Furthermore, our work can also serve as a plug-and-play module for other YOLO models. Typically, our method significantly advances the APs, APl, and AP of YOLOv8-N from 18%+, 52%+, and 37%+ to 20%+, 55%+, and 40%+, respectively, with even fewer parameters and MACs.

Previous deep learning-based event denoising methods mostly suffer from poor interpretability and difficulty in real-time processing due to their complex architecture designs. In this paper, we propose window-based event denoising, which simultaneously deals with a stack of events while existing element-based denoising focuses on one event each time. Besides, we give the theoretical analysis based on probability distributions in both temporal and spatial domains to improve interpretability. In temporal domain, we use timestamp deviations between processing events and central event to judge the temporal correlation and filter out temporal-irrelevant events. In spatial domain, we choose maximum a posteriori (MAP) to discriminate real-world event and noise and use the learned convolutional sparse coding to optimize the objective function. Based on the theoretical analysis, we build Temporal Window (TW) module and Soft Spatial Feature Embedding (SSFE) module to process temporal and spatial information separately, and construct a novel multi-scale window-based event denoising network, named WedNet. The high denoising accuracy and fast running speed of our WedNet enables us to achieve real-time denoising in complex scenes. Extensive experimental results verify the effectiveness and robustness of our WedNet. Our algorithm can remove event noise effectively and efficiently and improve the performance of downstream tasks.

We consider the problem of referring camouflaged object detection (Ref-COD), a new task that aims to segment specified camouflaged objects based on a small set of referring images with salient target objects. We first assemble a large-scale dataset, called R2C7K, which consists of 7 K images covering 64 object categories in real-world scenarios. Then, we develop a simple but strong dual-branch framework, dubbed R2CNet, with a reference branch embedding the common representations of target objects from referring images and a segmentation branch identifying and segmenting camouflaged objects under the guidance of the common representations. In particular, we design a Referring Mask Generation module to generate pixel-level prior mask and a Referring Feature Enrichment module to enhance the capability of identifying specified camouflaged objects. Extensive experiments show the superiority of our Ref-COD methods over their COD counterparts in segmenting specified camouflaged objects and identifying the main body of target objects.

Super-resolution (SR) has garnered significant attention within the computer vision community, driven by advances in deep learning (DL) techniques and the growing demand for high-quality visual applications. With the expansion of this field, numerous surveys have emerged. Most existing surveys focus on specific domains, lacking a comprehensive overview of this field. Here, we present an in-depth review of diverse SR methods, encompassing single-image SR (SISR), video SR (VSR), stereo SR (SSR), and light field SR (LFSR). We extensively cover over 150 SISR methods, nearly 70 VSR approaches, and approximately 30 techniques for SSR and LFSR. We analyze methodologies, datasets, evaluation protocols, empirical results, and complexity. In addition, we conducted a taxonomy based on each backbone structure according to the diverse purposes. We also explore valuable yet understudied open issues in the field. We believe that this work will serve as a valuable resource and offer guidance to researchers in this domain. To facilitate access to related work, we created a dedicated repository available at https://github.com/AVC2-UESTC/Holistic-Super-Resolution-Review

Class incremental learning (CIL) sequentially increases the number of classes, which often leads to catastrophic forgetting when fine-tuning on new classes. Existing approaches typically employ linear classifiers and expand them to accommodate new classes. However, conducting conventional classification inherently introduces feature drift in the image space upon the introduction of new classifiers, potentially disrupting the established distributions, and resulting in forgetting. In this paper, we propose a novel insight to reformulate the conventional classification as image-class matching (ICM) to mitigate the disruption. ICM independently encodes the image and the category and allows for the sharing of a matching classifier across all tasks, effectively stabilizing the feature space during the CIL process. To apply ICM to CIL, we introduce the Binary Matching Classification (BMC) framework, which employs cross attention to encode the matching relationship between images and each category to predict matching scores. When learning new tasks, BMC only requires the addition of category inputs without any structural changes. Moreover, we present a series of strategies to enhance the adaptation of BMC to CIL. Through simple regularization, our BMC framework achieves outstanding performance on various benchmarks including CIFAR-100, ImageNet-100, and ImageNet-1000 datasets. Our code is available athttps://github.com/Ethanhuhuhu/BMC.

We present a new boundary sensitive framework for polyp segmentation, termed Polyper.Our method is motivated by a clinical approach that seasoned medical practitioners often leverage the inherent features of interior polyp regions to tackle blurred boundaries.Inspired by this, we propose to explicitly leverages boundary regions to bolster the model's boundary discrimination capability while minimizing computational resource wastage. Our approach first extracts low-confidence boundary regions and high-confidence prediction regions from an initial segmentation map through differentiable morphological operators.Then, we design the boundary sensitive attention that concentrates on augmenting the features near the boundary regions using the high-confidence prediction region's characteristics to generate good segmentation results.Our proposed method can be seamlessly integrated with classical encoder networks, like ResNet-50, MiT-B1, and Swin Transformer.To evaludate the effectiveness of Polyper, we conduct experiments on five publicly available challenging datasets, and receive state-of-the-art performance on all of them. Code is available at https://github.com/haoshao-nku/medical_seg.git.

Traffic scene perception in computer vision is a critically important task to achieve intelligent cities. To date, most existing datasets focus on autonomous driving scenes. We observe that the models trained on those driving datasets often yield unsatisfactory results on traffic monitoring scenes. However, little effort has been put into improving the traffic monitoring scene understanding, mainly due to the lack of specific datasets. To fill this gap, we introduce a specialized traffic monitoring dataset, termed TSP6K, containing images from the traffic monitoring scenario, with high-quality pixel-level and instance-level annotations. The TSP6K dataset captures more crowded traffic scenes with several times more traffic participants than the existing driving scenes. We perform a detailed analysis of the dataset and comprehensively evaluate previous popular scene parsing methods, instance segmentation methods and unsupervised domain adaption methods. Furthermore, considering the vast difference in instance sizes, we propose a detail refining decoder for scene parsing, which recovers the details of different semantic regions in traffic scenes owing to the proposed TSP6K dataset. Experiments show its effectiveness in parsing the traffic monitoring scenes. Code and dataset are available at https://github.com/PengtaoJiang/TSP6K.

This paper presents a simple but performant semi-supervised semantic segmentation approach, called CorrMatch. Previous approaches mostly employ complicated training strategies to leverage unlabeled data but overlook the role of correlation maps in modeling the relationships between pairs of locations. We observe that the correlation maps not only enable clustering pixels of the same category easily but also contain good shape information, which previous works have omitted. Motivated by these, we aim to improve the use efficiency of unlabeled data by designing two novel label propagation strategies. First, we propose to conduct pixel propagation by modeling the pairwise similarities of pixels to spread the high-confidence pixels and dig out more. Then, we perform region propagation to enhance the pseudo labels with accurate class-agnostic masks extracted from the correlation maps. CorrMatch achieves great performance on popular segmentation benchmarks. Taking the DeepLabV3+ with ResNet-101 backbone as our segmentation model, we receive a 76%+ mIoU score on the Pascal VOC 2012 dataset with only 92 annotated images. Code is available at https://github.com/BBBBchan/CorrMatch.

Knowledge Distillation (KD) has been validated as an effective model compression technique for learning compact object detectors. Existing state-of-the-art KD methods for object detection are mostly based on feature imitation. In this paper, we present a general and effective prediction mimicking distillation scheme, called CrossKD, which delivers the intermediate features of the student's detection head to the teacher's detection head. The resulting crosshead predictions are then forced to mimic the teacher's predictions. This manner relieves the student's head from receiving contradictory supervision signals from the annotations and the teacher's predictions, greatly improving the student's detection performance. Moreover, as mimicking the teacher's predictions is the target of KD, CrossKD offers more task-oriented information in contrast with feature imitation. On MS COCO, with only prediction mimicking losses applied, our CrossKD boosts the average precision of GFL ResNet-50 with 1 × training schedule from 40.2 to 43.7, outperforming all existing KD methods. In addition, our method also works well when distilling detectors with heterogeneous backbones.

Previous multitask dense prediction methods based on the Mixture of Experts (MoE) have received great performance but they neglect the importance of explicitly modeling the global relations among all tasks. In this paper, we present a novel decoder-focused method for multitask dense prediction, called Mixture-of-Low-Rank-Experts (MLoRE). To model the global task relationships, MLoRE adds a generic convolution path to the original MoE structure, where each task feature can go through this path for explicit parameter sharing. Furthermore, to control the parameters and computational cost brought by the increase in the number of experts, we take inspiration from LoRA and propose to leverage the low-rank format of a vanilla con-volution in the expert network. Since the low-rank experts have fewer parameters and can be dynamically parameter-ized into the generic convolution, the parameters and computational cost do not change much with the increase of experts. Benefiting from this design, we increase the number of experts and its reception field to enlarge the representation capacity, facilitating multiple dense tasks learning in a unified network. Extensive experiments on the PASCAL-Context and NYUD-v2 benchmarks show that our MLoRE achieves superior performance compared to previous state-of-the-art methods on all metrics. Our code is available at https://github.com/YuqiYang213/MLoRE.

Recent progress in the text-driven 3D stylization of a single object has been considerably promoted by CLIP-based methods. However, the stylization of multi-object 3D scenes is still impeded in that the image-text pairs used for pre-training CLIP mostly consist of an object. Meanwhile, the local details of multiple objects may be susceptible to omission due to the existing supervision manner primarily relying on coarse-grained contrast of image-text pairs. To overcome these challenges, we present a novel framework, dubbed TeMO, to parse multi-object 3D scenes and edit their styles under the contrast supervision at multiple levels. We first propose a Decoupled Graph Attention (DGA) module to distinguishably reinforce the features of 3D surface points. Particularly, a cross-modal graph is constructed to align the object points accurately and noun phrases decoupled from the 3D mesh and textual description. Then, we develop a Cross-Grained Contrast (CGC) supervision system, where a fine-grained loss between the words in the textual description and the randomly rendered images are constructed to complement the coarse-grained loss. Extensive experiments show that our method can synthesize high-quality stylized content and outperform the existing methods over a wide range of multi-object 3D meshes.

The success of recent text-to-image diffusion models is largely due to their capacity to be guided by a complex text prompt, which enables users to precisely describe the desired content. However, these models struggle to effectively suppress the generation of undesired content, which is explicitly requested to be omitted from the generated image in the prompt. In this paper, we analyze how to manipulate the text embeddings and remove unwanted content from them. We introduce two contributions, which we refer to as soft-weighted regularization and inference-time text embedding optimization. The first regularizes the text embedding matrix and effectively suppresses the undesired content. The second method aims to further suppress the unwanted content generation of the prompt, and encourages the generation of desired content. We evaluate our method quantitatively and qualitatively on extensive experiments, validating its effectiveness. Furthermore, our method is generalizability to both the pixel-space diffusion models (i.e. DeepFloyd-IF) and the latent-space diffusion models (i.e. Stable Diffusion).

We present DFormer, a novel RGB-D pretraining framework to learn transferable representations for RGB-D segmentation tasks. DFormer has two new key innovations: 1) Unlike previous works that encode RGB-D information with RGB pretrained backbone, we pretrain the backbone using image-depth pairs from ImageNet-1K, and thus the DFormer is endowed with the capacity to encode RGB-D representations; 2) DFormer comprises a sequence of RGB-D blocks, which are tailored for encoding both RGB and depth information through a novel building block design. DFormer avoids the mismatched encoding of the 3D geometry relationships in depth maps by RGB pretrained backbones, which widely lies in existing methods but has not been resolved. We finetune the pretrained DFormer on two popular RGB-D tasks, i.e., RGB-D semantic segmentation and RGB-D salient object detection, with a lightweight decoder head. Experimental results show that our DFormer achieves new state-of-the-art performance on these two tasks with less than half of the computational cost of the current best methods on two RGB-D semantic segmentation datasets and five RGB-D salient object detection datasets. Code will be made publicly available.

Pre-trained vision-language models, e.g., CLIP, have been successfully applied to zero-shot semantic segmentation. Existing CLIP-based approaches primarily utilize visual features from the last layer to align with text embeddings, while they neglect the crucial information in intermediate layers that contain rich object details. However, we find that directly aggregating the multi-level visual features weakens the zero-shot ability for novel classes. The large differences between the visual features from different layers make these features hard to align well with the text embeddings. We resolve this problem by introducing a series of independent decoders to align the multi-level visual features with the text embeddings in a cascaded way, forming a novel but simple framework named Cascade-CLIP. Our Cascade-CLIP is flexible and can be easily applied to existing zero-shot semantic segmentation methods. Experimental results show that our simple Cascade-CLIP achieves superior zero-shot performance on segmentation benchmarks, like COCO-Stuff, Pascal-VOC, and Pascal-Context. Our code is available at https://github.com/HVision-NKU/Cascade-CLIP.

Synthetic Aperture Radar (SAR) object detection has gained significant attention recently due to its irreplaceable all-weather imaging capabilities. However, this research field suffers from both limited public datasets (mostly comprising <2K images with only mono-category objects) and inaccessible source code. To tackle these challenges, we establish a new benchmark dataset and an open-source method for large-scale SAR object detection. Our dataset, SARDet-100K, is a result of intense surveying, collecting, and standardizing 10 existing SAR detection datasets, providing a large-scale and diverse dataset for research purposes. To the best of our knowledge, SARDet-100K is the first COCO-level large-scale multi-class SAR object detection dataset ever created. With this high-quality dataset, we conducted comprehensive experiments and uncovered a crucial challenge in SAR object detection: the substantial disparities between the pretraining on RGB datasets and finetuning on SAR datasets in terms of both data domain and model structure. To bridge these gaps, we propose a novel Multi-Stage with Filter Augmentation (MSFA) pretraining framework that tackles the problems from the perspective of data input, domain transition, and model migration. The proposed MSFA method significantly enhances the performance of SAR object detection models while demonstrating exceptional generalizability and flexibility across diverse models. This work aims to pave the way for further advancements in SAR object detection. The dataset and code is available at \url{https://github.com/zcablii/SARDet_100K}.

Occupancy prediction, aiming at predicting the occupancy status within voxelized 3D environment, is quickly gaining momentum within the autonomous driving community. Mainstream occupancy prediction works first discretize the 3D environment into voxels, then perform classification on such dense grids. However, inspection on sample data reveals that the vast majority of voxels is unoccupied. Performing classification on these empty voxels demands suboptimal computation resource allocation, and reducing such empty voxels necessitates complex algorithm designs. To this end, we present a novel perspective on the occupancy prediction task: formulating it as a streamlined set prediction paradigm without the need for explicit space modeling or complex sparsification procedures. Our proposed framework, called OPUS, utilizes a transformer encoder-decoder architecture to simultaneously predict occupied locations and classes using a set of learnable queries. Firstly, we employ the Chamfer distance loss to scale the set-to-set comparison problem to unprecedented magnitudes, making training such model end-to-end a reality. Subsequently, semantic classes are adaptively assigned using nearest neighbor search based on the learned locations. In addition, OPUS incorporates a suite of non-trivial strategies to enhance model performance, including coarse-to-fine learning, consistent point sampling, and adaptive re-weighting, etc. Finally, compared with current state-of-the-art methods, our lightest model achieves superior RayIoU on the Occ3D-nuScenes dataset at near 2x FPS, while our heaviest model surpasses previous best results by 6.1 RayIoU.

For recent diffusion-based generative models, maintaining consistent content across a series of generated images, especially those containing subjects and complex details, presents a significant challenge. In this paper, we propose a simple but effective self-attention mechanism, termed Consistent Self-Attention, that boosts the consistency between the generated images. It can be used to augment pre-trained diffusion-based text-to-image models in a zero-shot manner. Based on the images with consistent content, we further show that our method can be extended to long range video generation by introducing a semantic space temporal motion prediction module, named Semantic Motion Predictor. It is trained to estimate the motion conditions between two provided images in the semantic spaces. This module converts the generated sequence of images into videos with smooth transitions and consistent subjects that are more stable than the modules based on latent spaces only, especially in the context of long video generation. By merging these two novel components, our framework, referred to as StoryDiffusion, can describe a text-based story with consistent images or videos encompassing a rich variety of contents. The proposed StoryDiffusion encompasses pioneering explorations in visual story generation with the presentation of images and videos, which we hope could inspire more research from the aspect of architectural modifications.

Vision Transformers have been the most popular network architecture in visual recognition recently due to the strong ability of encode global information. However, its high computational cost when processing high-resolution images limits the applications in downstream tasks. In this paper, we take a deep look at the internal structure of self-attention and present a simple Transformer style convolutional neural network (ConvNet) for visual recognition. By comparing the design principles of the recent ConvNets and Vision Transformers, we propose to simplify the self-attention by leveraging a convolutional modulation operation. We show that such a simple approach can better take advantage of the large kernels ( ≥ 7×7) nested in convolutional layers and we observe a consistent performance improvement when gradually increasing the kernel size from 5×5 to 21×21. We build a family of hierarchical ConvNets using the proposed convolutional modulation, termed Conv2Former. Our network is simple and easy to follow. Experiments show that our Conv2Former outperforms existent popular ConvNets and vision Transformers, like Swin Transformer and ConvNeXt in all ImageNet classification, COCO object detection and ADE20 k semantic segmentation.

Masked image modeling (MIM) has achieved promising results on various vision tasks. However, the limited discriminability of learned representation manifests there is still plenty to go for making a stronger vision learner. Towards this goal, we propose Contrastive Masked Autoencoders (CMAE), a new self-supervised pre-training method for learning more comprehensive and capable vision representations. By elaboratively unifying contrastive learning (CL) and masked image model (MIM) through novel designs, CMAE leverages their respective advantages and learns representations with both strong instance discriminability and local perceptibility. Specifically, CMAE consists of two branches where the online branch is an asymmetric encoder-decoder and the momentum branch is a momentum updated encoder. During training, the online encoder reconstructs original images from latent representations of masked images to learn holistic features. The momentum encoder, fed with the full images, enhances the feature discriminability via contrastive learning with its online counterpart. To make CL compatible with MIM, CMAE introduces two new components, i.e., pixel shifting for generating plausible positive views and feature decoder for complementing features of contrastive pairs. Thanks to these novel designs, CMAE effectively improves the representation quality and transfer performance over its MIM counterpart. CMAE achieves the state-of-the-art performance on highly competitive benchmarks of image classification, semantic segmentation and object detection. Notably, CMAE-Base achieves 85.3% top-1 accuracy on ImageNet and 52.5% mIoU on ADE20k, surpassing previous best results by 0.7% and 1.8% respectively.

How to identify and segment camouflaged objects from the background is challenging. Inspired by the multi-head self-attention in Transformers, we present a simple masked separable attention (MSA) for camouflaged object detection. We first separate the multi-head self-attention into three parts, which are responsible for distinguishing the camouflaged objects from the background using different mask strategies. Furthermore, we propose to capture high-resolution semantic representations progressively based on a simple top-down decoder with the proposed MSA to attain precise segmentation results. These structures plus a backbone encoder form a new model, dubbed CamoFormer. Extensive experiments show that CamoFormer achieves new state-of-the-art performance on three widely-used camouflaged object detection benchmarks. To better evaluate the performance of the proposed CamoFormer around the border regions, we propose to use two new metrics, i.e., BR-M and BR-F. There are on average ∼ 5% relative improvements over previous methods in terms of S-measure and weighted F-measure.

A fundamental limitation of object detectors is that they suffer from "spatial bias", and in particular perform less satisfactorily when detecting objects near image borders. For a long time, there has been a lack of effective ways to measure and identify spatial bias, and little is known about where it comes from and what degree it is. To this end, we present a new zone evaluation protocol, extending from the traditional evaluation to a more generalized one, which measures the detection performance over zones, yielding a series of Zone Precisions (ZPs). For the first time, we provide numerical results, showing that the object detectors perform quite unevenly across the zones. Surprisingly, the detector's performance in the 96% border zone of the image does not reach the AP value (Average Precision, commonly regarded as the average detection performance in the entire image zone). To better understand spatial bias, a series of heuristic experiments are conducted. Our investigation excludes two intuitive conjectures about spatial bias that the object scale and the absolute positions of objects barely influence the spatial bias. We find that the key lies in the human-imperceptible divergence in data patterns between objects in different zones, thus eventually forming a visible performance gap between the zones. With these findings, we finally discuss a future direction for object detection, namely, spatial disequilibrium problem, aiming at pursuing a balanced detection ability over the entire image zone. By broadly evaluating 10 popular object detectors and 5 detection datasets, we shed light on the spatial bias of object detectors. We hope this work could raise a focus on detection robustness.

We present All-Pairs Multi-Field Transforms (AMT), a new network architecture for video frame interpolation. It is based on two essential designs. First, we build bidirectional correlation volumes for all pairs of pixels, and use the predicted bilateral flows to retrieve correlations for updating both flows and the interpolated content feature. Second, we derive multiple groups of fine-grained flow fields from one pair of updated coarse flows for performing backward warping on the input frames separately. Combining these two designs enables us to generate promising task-oriented flows and reduce the difficulties in modeling large motions and handling occluded areas during frame interpolation. These qualities promote our model to achieve state-of-the-art performance on various benchmarks with high efficiency. Moreover, our convolution-based model competes favorably compared to Transformer-based models in terms of accuracy and efficiency. Our code is available at https://github.com/MCG-NKU/AMT.

Recent research on remote sensing object detection has largely focused on improving the representation of oriented bounding boxes but has overlooked the unique prior knowledge presented in remote sensing scenarios. Such prior knowledge can be useful because tiny remote sensing objects may be mistakenly detected without referencing a sufficiently long-range context, which can vary for different objects. This paper considers these priors and proposes the lightweight Large Selective Kernel Network (LSKNet). LSKNet can dynamically adjust its large spatial receptive field to better model the ranging context of various objects in remote sensing scenarios. To our knowledge, large and selective kernel mechanisms have not been previously explored in remote sensing object detection. Without bells and whistles, our lightweight LSKNet sets new state-of-the-art scores on standard benchmarks, i.e., HRSC2016 (98.46% mAP), DOTA-v1.0 (81.85% mAP), and FAIR1M-v1.0 (47.87% mAP).

Previous works have shown that increasing the window size for Transformer-based image super-resolution models (e.g., SwinIR) can significantly improve the model performance but the computation overhead is also considerable. In this paper, we present SRFormer, a simple but novel method that can enjoy the benefit of large window self-attention but introduces even less computational burden. The core of our SRFormer is the permuted self-attention (PSA), which strikes an appropriate balance between the channel and spatial information for self-attention. Our PSA is simple and can be easily applied to existing super-resolution networks based on window self-attention. Without any bells and whistles, we show that our SRFormer achieves a 33.86dB PSNR score on the Urban100 dataset, which is 0.46dB higher than that of SwinIR but uses fewer parameters and computations. We hope our simple and effective approach can serve as a useful tool for future research in super-resolution model design. Our code is available at https://github.com/HVision-NKU/SRFormer.

We explore the potential of pooling techniques on the task of salient object detection by expanding its role in convolutional neural networks. In general, two pooling-based modules are proposed. A global guidance module (GGM) is first built based on the bottom-up pathway of the U-shape architecture, which aims to guide the location information of the potential salient objects into layers at different feature levels. A feature aggregation module (FAM) is further designed to seamlessly fuse the coarse-level semantic information with the fine-level features in the top-down pathway. We can progressively refine the high-level semantic features with these two modules and obtain detail enriched saliency maps. Experimental results show that our proposed approach can locate the salient objects more accurately with sharpened details and substantially improve the performance compared with the existing state-of-the-art methods. Besides, our approach is fast and can run at a speed of 53 FPS when processing a 300 ×400 image. To make our approach better applied to mobile applications, we take MobileNetV2 as our backbone and re-tailor the structure of our pooling-based modules. Our mobile version model achieves a running speed of 66 FPS yet still performs better than most existing state-of-the-art methods. To verify the generalization ability of the proposed method, we apply it to the edge detection, RGB-D salient object detection, and camouflaged object detection tasks, and our method achieves better results than the corresponding state-of-the-art methods of these three tasks. Code can be found at http://mmcheng.net/poolnet/.

In this paper, we present Vision Permutator, a conceptually simple and data efficient MLP-like architecture for visual recognition. By realizing the importance of the positional information carried by 2D feature representations, unlike recent MLP-like models that encode the spatial information along the flattened spatial dimensions, Vision Permutator separately encodes the feature representations along the height and width dimensions with linear projections. This allows Vision Permutator to capture long-range dependencies and meanwhile avoid the attention building process in transformers. The outputs are then aggregated in a mutually complementing manner to form expressive representations. We show that our Vision Permutators are formidable competitors to convolutional neural networks (CNNs) and vision transformers. Without the dependence on spatial convolutions or attention mechanisms, Vision Permutator achieves 81.5% top-1 accuracy on ImageNet without extra large-scale training data (e.g., ImageNet-22k) using only 25M learnable parameters, which is much better than most CNNs and vision transformers under the same model size constraint. When scaling up to 88M, it attains 83.2% top-1 accuracy, greatly improving the performance of recent state-of-the-art MLP-like networks for visual recognition. We hope this work could encourage research on rethinking the way of encoding spatial information and facilitate the development of MLP-like models. PyTorch/MindSpore/Jittor code is available at https://github.com/Andrew-Qibin/VisionPermutator.

Recently, Vision Transformers (ViTs) have been broadly explored in visual recognition. With low efficiency in encoding fine-level features, the performance of ViTs is still inferior to the state-of-the-art CNNs when trained from scratch on a midsize dataset like ImageNet. Through experimental analysis, we find it is because of two reasons: 1) the simple tokenization of input images fails to model the important local structure such as edges and lines, leading to low training sample efficiency; 2) the redundant attention backbone design of ViTs leads to limited feature richness for fixed computation budgets and limited training samples. To overcome such limitations, we present a new simple and generic architecture, termed Vision Outlooker (VOLO), which implements a novel outlook attention operation that dynamically conduct the local feature aggregation mechanism in a sliding window manner across the input image. Unlike self-attention that focuses on modeling global dependencies of local features at a coarse level, our outlook attention targets at encoding finer-level features, which is critical for recognition but ignored by self-attention. Outlook attention breaks the bottleneck of self-attention whose computation cost scales quadratically with the input spatial dimension, and thus is much more memory efficient. Compared to our Tokens-To-Token Vision Transformer (T2T-ViT), VOLO can more efficiently encode fine-level features that are essential for high-performance visual recognition. Experiments show that with only 26.6 M learnable parameters, VOLO achieves 84.2% top-1 accuracy on ImageNet-1 K without using extra training data, 2.7% better than T2T-ViT with a comparable number of parameters. When the model size is scaled up to 296 M parameters, its performance can be further improved to 87.1%, setting a new record for ImageNet-1 K classification. In addition, we also take the proposed VOLO as pretrained models and report superior performance on downstream tasks, such as semantic segmentation. Code is available at https://github.com/sail-sg/volo.

Previous knowledge distillation (KD) methods for object detection mostly focus on feature imitation instead of mimicking the prediction logits due to its inefficiency in distilling the localization information. In this paper, we investigate whether logit mimicking always lags behind feature imitation. Towards this goal, we first present a novel localization distillation (LD) method which can efficiently transfer the localization knowledge from the teacher to the student. Second, we introduce the concept of valuable localization region that can aid to selectively distill the classification and localization knowledge for a certain region. Combining these two new components, for the first time, we show that logit mimicking can outperform feature imitation and the absence of localization distillation is a critical reason for why logit mimicking under-performs for years. The thorough studies exhibit the great potential of logit mimicking that can significantly alleviate the localization ambiguity, learn robust feature representation, and ease the training difficulty in the early stage. We also provide the theoretical connection between the proposed LD and the classification KD, that they share the equivalent optimization effect. Our distillation scheme is simple as well as effective and can be easily applied to both dense horizontal object detectors and rotated object detectors. Extensive experiments on the MS COCO, PASCAL VOC, and DOTA benchmarks demonstrate that our method can achieve considerable AP improvement without any sacrifice on the inference speed. Our source code and pretrained models are publicly available at https://github.com/HikariTJU/LD.

Vision transformers have recently attained state-of-the-art results in visual recognition tasks. Their success is largely attributed to the self-attention component, which models the global dependencies among the image patches (tokens) and aggregates them into higher-level features. However, self-attention brings significant training difficulties to ViTs. Many recent works thus develop various new self-attention components to alleviate this issue. In this article, instead of developing complicated self-attention mechanism, we aim to explore simple approaches to fully release the potential of the vanilla self-attention. We first study the token selection behavior of self-attention and find that it suffers from a low diversity due to attention over-smoothing, which severely limits its effectiveness in learning discriminative token features. We then develop simple approaches to enhance selectivity and diversity for self-attention in token selection. The resulted token selector module can server as a drop-in module for various ViT backbones and consistently boost their performance. Significantly, they enable ViTs to achieve 84.6% top-1 classification accuracy on ImageNet with only 25M parameters. When scaled up to 81M parameters, the result can be further improved to 86.1%. In addition, we also present comprehensive experiments to demonstrate the token selector can be applied to a variety of transformer-based models to boost their performance for image classification, semantic segmentation and NLP tasks. Code is available at https://github.com/zhoudaquan/dvit_repo.

Mining precise class-aware attention maps, a.k.a, class activation maps, is essential for weakly supervised semantic segmentation. In this paper, we present L2G, a simple online local-to-global knowledge transfer framework for high-quality object attention mining. We observe that classification models can discover object regions with more details when replacing the input image with its local patches. Taking this into account, we first leverage a local classification network to extract attentions from multiple local patches randomly cropped from the input image. Then, we utilize a global network to learn complementary attention knowledge across multiple local attention maps online. Our framework conducts the global network to learn the captured rich object detail knowledge from a global view and thereby produces high-quality attention maps that can be directly used as pseudo annotations for semantic segmentation networks. Experiments show that our method attains 72.1% and 44.2% mIoU scores on the validation set of PASCAL VOC 2012 and MS COCO 2014, respectively, setting new state-of-the-art records. Code is available at https://github.com/PengtaoJiang/L2G.

Knowledge distillation (KD) has witnessed its powerful capability in learning compact models in object detection. Previous KD methods for object detection mostly focus on imitating deep features within the imitation regions instead of mimicking classification logit due to its inefficiency in distilling localization information and trivial improvement. In this paper, by reformulating the knowledge distillation process on localization, we present a novel localization distillation (LD) method which can efficiently transfer the localization knowledge from the teacher to the student. Moreover, we also heuristically introduce the concept of valuable localization region that can aid to selectively distill the semantic and localization knowledge for a certain region. Combining these two new components, for the first time, we show that logit mimicking can outperform feature imitation and localization knowledge distillation is more important and efficient than semantic knowledge for distilling object detectors. Our distillation scheme is simple as well as effective and can be easily applied to different dense object detectors. Experiments show that our LD can boost the AP score of GFocal-ResNet-50 with a single-scale 1 x training schedule from 40.1 to 42.1 on the COCO benchmark without any sacrifice on the inference speed. Our source code and pretrained models are publicly available at https://github.com/HikariTJU/LD.

High-dynamic-range (HDR) media resources that preserve high contrast and more details in shadow and highlight areas in television are becoming increasingly popular for modern display technology compared to the widely available standard-dynamic-range (SDR) media resources. However, due to the exorbitant price of HDR cameras, researchers have attempted to develop the SDR-to-HDR techniques to convert the abundant SDR media resources to the HDR versions for cost-saving. Recent SDR-to-HDR methods mostly apply the image-adaptive modulation scheme to dynamically modulate the local contrast. However, these methods often fail to properly capture the low-frequency cues, resulting in artifacts in the low-frequency regions and low visual quality. Motivated by the Discrete Cosine Transform (DCT), in this paper, we propose a Frequency-aware Modulation Network (FMNet) to enhance the contrast in a frequency-adaptive way for SDR-to-HDR translation. Specifically, we design a frequency-aware modulation block that can dynamically modulate the features according to its frequency-domain responses. This allows us to reduce the structural distortions and artifacts in the translated low-frequency regions and reconstruct high-quality HDR content in the translated results. Experimental results on the HDRTV1K dataset show that our FMNet outperforms previous methods and the perceptual quality of the generated HDR images can be largely improved. Our code is available at https://github.com/MCG-NKU/FMNet.

We present SegNeXt, a simple convolutional network architecture for semantic segmentation. Recent transformer-based models have dominated the field of se- mantic segmentation due to the efficiency of self-attention in encoding spatial information. In this paper, we show that convolutional attention is a more efficient and effective way to encode contextual information than the self-attention mech- anism in transformers. By re-examining the characteristics owned by successful segmentation models, we discover several key components leading to the perfor- mance improvement of segmentation models. This motivates us to design a novel convolutional attention network that uses cheap convolutional operations. Without bells and whistles, our SegNeXt significantly improves the performance of previous state-of-the-art methods on popular benchmarks, including ADE20K, Cityscapes, COCO-Stuff, Pascal VOC, Pascal Context, and iSAID. Notably, SegNeXt out- performs EfficientNet-L2 w/ NAS-FPN and achieves 90.6% mIoU on the Pascal VOC 2012 test leaderboard using only 1/10 parameters of it. On average, SegNeXt achieves about 2.0% mIoU improvements compared to the state-of-the-art methods on the ADE20K datasets with the same or fewer computations.

Object attention maps generated by image classifiers are usually used as priors for weakly supervised semantic segmentation. However, attention maps usually locate the most discriminative object parts. The lack of integral object localization maps heavily limits the performance of weakly supervised segmentation approaches. This paper attempts to investigate a novel way to identify entire object regions in a weakly supervised manner. We observe that image classifiers' attention maps at different training phases may focus on different parts of the target objects. Based on this observation, we propose an online attention accumulation (OAA) strategy that utilizes the attention maps at different training phases to obtain more integral object regions. Specifically, we maintain a cumulative attention map for each target category in each training image and utilize it to record the discovered object regions at different training phases. Albeit OAA can effectively mine more object regions for most images, for some training images, the range of the attention movement is not large, limiting the generation of integral object attention regions. To overcome this problem, we propose incorporating an attention drop layer into the online attention accumulation process to enlarge the range of attention movement during training explicitly. Our method (OAA) can be plugged into any classification network and progressively accumulate the discriminative regions into cumulative attention maps as the training process goes. Additionally, we also explore utilizing the final cumulative attention maps to serve as the pixel-level supervision, which can further assist the network in discovering more integral object regions. When applying the resulting attention maps to the weakly supervised semantic segmentation task, our approach improves the existing state-of-the-art methods on the PASCAL VOC 2012 segmentation benchmark, achieving a mIoU score of 67.2 percent on the test set.

Attention mechanisms have been proven highly effective for salient object detection. Most previous works utilize attention as a self-gated module to reweigh the feature maps at different levels independently. However, they are limited to certain-level guidance and could not satisfy the need of both accurately detecting intact objects and maintaining their detailed boundaries. In this paper, we build dense attention upon features from multiple levels simultaneously and propose a novel Dense Attentive Feature Enhancement (DAFE) module for efficient feature enhancement in saliency detection. DAFE stacks several attentional units and densely connects attentive feature output from current unit to its all subsequent units. This allows feature maps at deep units to absorb attentive information from shallow units, thus more discriminative information can be efficiently selected at the final output. Note that DAFE is plug and play, which can be effortlessly inserted into any saliency or video saliency models for their performance improvements. We further instantiate a highly effective Dense Attentive Feature Enhancement Network (DAFE-Net) for accurate salient object detection. DAFE-Net constructs DAFE over the aggregation feature that contains both semantics and saliency details, the entire salient objects and their boundaries can be well retained through dense attentions. Extensive experiments demonstrate that the proposed DAFE module is highly effective, and the DAFE-Net performs favorably compared with state-of-the-art approaches.

Recent studies on mobile network design have demonstrated the remarkable effectiveness of channel attention (e.g., the Squeeze-and-Excitation attention) for lifting model performance, but they generally neglect the positional information, which is important for generating spatially selective attention maps. In this paper, we propose a novel attention mechanism for mobile networks by embedding positional information into channel attention, which we call "coordinate attention". Unlike channel attention that transforms a feature tensor to a single feature vector via 2D global pooling, the coordinate attention factorizes channel attention into two 1D feature encoding processes that aggregate features along the two spatial directions, respectively. In this way, long-range dependencies can be captured along one spatial direction and meanwhile precise positional information can be preserved along the other spatial direction. The resulting feature maps are then encoded separately into a pair of direction-aware and position-sensitive attention maps that can be complementarily applied to the input feature map to augment the representations of the objects of interest. Our coordinate attention is simple and can be flexibly plugged into classic mobile networks, such as MobileNetV2, MobileNeXt, and EfficientNet with nearly no computational overhead. Extensive experiments demonstrate that our coordinate attention is not only beneficial to ImageNet classification but more interestingly, behaves better in down-stream tasks, such as object detection and semantic segmentation. Code is available at https://github.com/Andrew-Qibin/CoordAttention.

Current neural architecture search (NAS) algorithms still require expert knowledge and effort to design a search space for network construction. In this paper, we consider automating the search space design to minimize human interference, which however faces two challenges: the ex-plosive complexity of the exploration space and the expensive computation cost to evaluate the quality of different search spaces. To solve them, we propose a novel differentiable evolutionary framework named AutoSpace, which evolves the search space to an optimal one with following novel techniques: a differentiable fitness scoring function to efficiently evaluate the performance of cells and a reference architecture to speedup the evolution procedure and avoid falling into sub-optimal solutions. The frame-work is generic and compatible with additional computational constraints, making it feasible to learn specialized search spaces that fit different computational bud-gets. With the learned search space, the performance of recent NAS algorithms can be improved significantly compared with using previously manually designed spaces. Remarkably, the models generated from the new search space achieve 77.8% top-1 accuracy on ImageNet under the mobile setting (MAdds 500M), outperforming previous SOTA EfficientNet-B0 by≤0.7%. https://github.com/zhoudaquan/AutoSpace.git.

In this paper, we present token labeling---a new training objective for training high-performance vision transformers (ViTs). Different from the standard training objective of ViTs that computes the classification loss on an additional trainable class token, our proposed one takes advantage of all the image patch tokens to compute the training loss in a dense manner. Specifically, token labeling reformulates the image classification problem into multiple token-level recognition problems and assigns each patch token with an individual location-specific supervision generated by a machine annotator. Experiments show that token labeling can clearly and consistently improve the performance of various ViT models across a wide spectrum. For a vision transformer with 26M learnable parameters serving as an example, with token labeling, the model can achieve 84.4% Top-1 accuracy on ImageNet. The result can be further increased to 86.4% by slightly scaling the model size up to 150M, delivering the minimal-sized model among previous models (250M+) reaching 86%. We also show that token labeling can clearly improve the generalization capability of the pretrained models on downstream tasks with dense prediction, such as semantic segmentation. Our code and model are publiclyavailable at https://github.com/zihangJiang/TokenLabeling.

Benefiting from the capability of building interdependencies among channels or spatial locations, attention mechanisms have been extensively studied and broadly used in a variety of computer vision tasks recently. In this paper, we investigate light-weight but effective attention mechanisms and present triplet attention, a novel method for computing attention weights by capturing crossdimension interaction using a three-branch structure. For an input tensor, triplet attention builds inter-dimensional dependencies by the rotation operation followed by residual transformations and encodes inter-channel and spatial information with negligible computational overhead. Our method is simple as well as efficient and can be easily plugged into classic backbone networks as an add-on module. We demonstrate the effectiveness of our method on various challenging tasks including image classification on ImageNet-1k and object detection on MSCOCO and PASCAL VOC datasets. Furthermore, we provide extensive insight into the performance of triplet attention by visually inspecting the GradCAM and GradCAM++ results. The empirical evaluation of our method supports our intuition on the importance of capturing dependencies across dimensions when computing attention weights. Code for this paper can be publicly accessed at https://github.com/LandskapeAI/triplet-attention.

Domain adversarial training has become a prevailing and effective paradigm for unsupervised domain adaptation (UDA). To successfully align the multi-modal data structures across domains, the following works exploit discriminative information in the adversarial training process, e.g., using multiple class-wise discriminators and involving conditional information in the input or output of the domain discriminator. However, these methods either require non-trivial model designs or are inefficient for UDA tasks. In this work, we attempt to address this dilemma by devising simple and compact conditional domain adversarial training methods. We first revisit the simple concatenation conditioning strategy where features are concatenated with output predictions as the input of the discriminator. We find the concatenation strategy suffers from the weak conditioning strength. We further demonstrate that enlarging the norm of concatenated predictions can effectively energize the conditional domain alignment. Thus we improve concatenation conditioning by normalizing the output predictions to have the same norm of features, and term the derived method as Normalized OutpUt coNditioner (NOUN). However, conditioning on raw output predictions for domain alignment, NOUN suffers from inaccurate predictions of the target domain. To this end, we propose to condition the cross-domain feature alignment in the prototype space rather than in the output space. Combining the novel prototype-based conditioning with NOUN, we term the enhanced method as PROtotype-based Normalized OutpUt coNditioner (PRONOUN). Experiments on both object recognition and semantic segmentation show that NOUN can effectively align the multi-modal structures across domains and even outperform state-of-the-art domain adversarial training methods. Together with prototype-based conditioning, PRONOUN further improves the adaptation performance over NOUN on multiple object recognition benchmarks for UDA. Code is available at https://github.com/tim-learn/NOUN.

The class activation maps are generated from the final convolutional layer of CNN. They can highlight discriminative object regions for the class of interest. These discovered object regions have been widely used for weakly-supervised tasks. However, due to the small spatial resolution of the final convolutional layer, such class activation maps often locate coarse regions of the target objects, limiting the performance of weakly-supervised tasks that need pixel-accurate object locations. Thus, we aim to generate more fine-grained object localization information from the class activation maps to locate the target objects more accurately. In this paper, by rethinking the relationships between the feature maps and their corresponding gradients, we propose a simple yet effective method, called LayerCAM. It can produce reliable class activation maps for different layers of CNN. This property enables us to collect object localization information from coarse (rough spatial localization) to fine (precise fine-grained details) levels. We further integrate them into a high-quality class activation map, where the object-related pixels can be better highlighted. To evaluate the quality of the class activation maps produced by LayerCAM, we apply them to weakly-supervised object localization and semantic segmentation. Experiments demonstrate that the class activation maps generated by our method are more effective and reliable than those by the existing attention methods. The code will be made publicly available.

Feature pyramid network (FPN) based models, which fuse the semantics and salient details in a progressive manner, have been proven highly effective in salient object detection. However, it is observed that these models often generate saliency maps with incomplete object structures or unclear object boundaries, due to the indirect information propagation among distant layers that makes such fusion structure less effective. In this work, we propose a novel Cross-layer Feature Pyramid Network (CFPN), in which direct cross-layer communication is enabled to improve the progressive fusion in salient object detection. Specifically, the proposed network first aggregates multi-scale features from different layers into feature maps that have access to both the high- and low- level information. Then, it distributes the aggregated features to all the involved layers to gain access to richer context. In this way, the distributed features per layer own both semantics and salient details from all other layers simultaneously, and suffer reduced loss of important information during the progressive feature fusion. At last, CFPN fuses the distributed features of each layer stage-by-stage. This way, the high-level features that contain context useful for locating complete objects are preserved until the final output layer, and the low-level features that contain spatial structure details are embedded into each layer to preserve spatial structural details. Extensive experimental results over six widely used salient object detection benchmarks and with three popular backbones clearly demonstrate that CFPN can accurately locate fairly complete salient regions and effectively segment the object boundaries.

Label smoothing is an effective regularization tool for deep neural networks (DNNs), which generates soft labels by applying a weighted average between the uniform distribution and the hard label. It is often used to reduce the overfitting problem of training DNNs and further improve classification performance. In this paper, we aim to investigate how to generate more reliable soft labels. We present an Online Label Smoothing (OLS) strategy, which generates soft labels based on the statistics of the model prediction for the target category. The proposed OLS constructs a more reasonable probability distribution between the target categories and non-target categories to supervise DNNs. Experiments demonstrate that based on the same classification models, the proposed approach can effectively improve the classification performance on CIFAR-100, ImageNet, and fine-grained datasets. Additionally, the proposed method can significantly improve the robustness of DNN models to noisy labels compared to current label smoothing approaches. The source code is available at our project page: https://mmcheng.net/ols/.

Spatial pooling has been proven highly effective to capture long-range contextual information for pixel-wise prediction tasks, such as scene parsing. In this paper, beyond conventional spatial pooling that usually has a regular shape of NxN, we rethink the formulation of spatial pooling by introducing a new pooling strategy, called strip pooling, which considers a long but narrow kernel, i.e., 1xN or Nx1. Based on strip pooling, we further investigate spatial pooling architecture design by 1) introducing a new strip pooling module that enables backbone networks to efficiently model long-range dependencies; 2) presenting a novel building block with diverse spatial pooling as a core; and 3) systematically comparing the performance of the proposed strip pooling and conventional spatial pooling techniques. Both novel pooling-based designs are lightweight and can serve as an efficient plug-and-play modules in existing scene parsing networks. Extensive experiments on Cityscapes and ADE20K benchmarks demonstrate that our simple approach establishes new state-of-the-art results. Code is available at https://github.com/Andrew-Qibin/SPNet.

Recent advances on CNNs are mostly devoted to designing more complex architectures to enhance their representation learning capacity. In this paper, we consider how to improve the basic convolutional feature transformation process of CNNs without tuning the model architectures. To this end, we present a novel self-calibrated convolutions that explicitly expand fields-of-view of each convolutional layers through internal communications and hence enrich the output features. In particular, unlike the standard convolutions that fuse spatial and channel-wise information using small kernels (e.g., 3×3), self-calibrated convolutions adaptively build long-range spatial and inter-channel dependencies around each spatial location through a novel self-calibration operation. Thus, it can help CNNs generate more discriminative representations by explicitly incorporating richer information. Our self-calibrated convolution design is simple and generic, and can be easily applied to augment standard convolutional layers without introducing extra parameters and complexity. Extensive experiments demonstrate that when applying self-calibrated convolutions into different backbones, our networks can significantly improve the baseline models in a variety of vision tasks, including image recognition, object detection, instance segmentation, and keypoint detection, with no need to change the network architectures. We hope this work could provide a promising way for future research in designing novel convolutional feature transformations for improving convolutional networks. Code is available on the project page.

In this paper, we consider salient instance segmentation. As well as producing bounding boxes, our network also outputs high-quality instance-level segments as initial selections to indicate the regions of interest. Taking into account the category-independent property of each target, we design a single stage salient instance segmentation framework, with a novel segmentation branch. Our new branch regards not only local context inside each detection window but also the surrounding context, enabling us to distinguish instances in the same scope even with partial occlusion. Our network is end-to-end trainable and is fast (running at 40 fps for images with resolution 320 × 320). We evaluate our approach on a publicly available benchmark and show that it outperforms alternative solutions. We also provide a thorough analysis of our design choices to help readers better understand the function of each part of our network. Source code can be found at https://github.com/RuochenFan/S4Net.

Salient object segmentation, edge detection, and skeleton extraction are three contrasting low-level pixel-wise vision problems, where existing works mostly focused on designing tailored methods for each individual task. However, it is inconvenient and inefficient to store a pre-trained model for each task and perform multiple different tasks in sequence. There are methods that solve specific related tasks jointly but require datasets with different types of annotations supported at the same time. In this paper, we first show some similarities shared by these tasks and then demonstrate how they can be leveraged for developing a unified framework that can be trained end-to-end. In particular, we introduce a selective integration module that allows each task to dynamically choose features at different levels from the shared backbone based on its own characteristics. Furthermore, we design a task-adaptive attention module, aiming at intelligently allocating information for different tasks according to the image content priors. To evaluate the performance of our proposed network on these tasks, we conduct exhaustive experiments on multiple representative datasets. We will show that though these tasks are naturally quite different, our network can work well on all of them and even perform better than current single-purpose state-of-the-art methods. In addition, we also conduct adequate ablation analyses that provide a full understanding of the design principles of the proposed framework. To facilitate future research, source code will be released.

We consider an interesting problem---salient instance segmentation. Other than producing approximate bounding boxes, our network also outputs high-quality instance-level segments. Taking into account the category-independent property of each target, we design a single stage salient instance segmentation framework, with a novel segmentation branch. Our new branch regards not only local context inside each detection window but also its surrounding context, enabling us to distinguish the instances in the same scope even with obstruction. Our network is end-to-end trainable and runs at a fast speed (40 fps when processing an image with resolution 320 x 320). We evaluate our approach on a public available benchmark and show that it outperforms other alternative solutions. We also provide a thorough analysis of the design choices to help readers better understand the functions of each part of our network. The source code can be found at https://github.com/RuochenFan/S4Net.

We solve the problem of salient object detection by investigating how to expand the role of pooling in convolutional neural networks. Based on the U-shape architecture, we first build a global guidance module (GGM) upon the bottom-up pathway, aiming at providing layers at different feature levels the location information of potential salient objects. We further design a feature aggregation module (FAM) to make the coarse-level semantic information well fused with the fine-level features from the top-down path- way. By adding FAMs after the fusion operations in the top-down pathway, coarse-level features from the GGM can be seamlessly merged with features at various scales. These two pooling-based modules allow the high-level semantic features to be progressively refined, yielding detail enriched saliency maps. Experiment results show that our proposed approach can more accurately locate the salient objects with sharpened details and hence substantially improve the performance compared to the previous state-of-the-arts. Our approach is fast as well and can run at a speed of more than 30 FPS when processing a 300×400 image. Code can be found at http://mmcheng.net/poolnet/.

Object attention maps generated by image classifiers are usually used as priors for weakly-supervised segmentation approaches. However, normal image classifiers produce attention only at the most discriminative object parts, which limits the performance of weakly-supervised segmentation task. Therefore, how to effectively identify entire object regions in a weakly-supervised manner has always been a challenging and meaningful problem. We observe that the attention maps produced by a classification network continuously focus on different object parts during training. In order to accumulate the discovered different object parts, we propose an online attention accumulation (OAA) strategy which maintains a cumulative attention map for each target category in each training image so that the integral object regions can be gradually promoted as the training goes. These cumulative attention maps, in turn, serve as the pixel-level supervision, which can further assist the network in discovering more integral object regions. Our method (OAA) can be plugged into any classification network and progressively accumulate the discriminative regions into integral objects as the training process goes. Despite its simplicity, when applying the resulting attention maps to the weakly-supervised semantic segmentation task, our approach improves the existing state-of-the-art methods on the PASCAL VOC 2012 segmentation benchmark, achieving a mIoU score of 66.4% on the test set. Code is available at https://mmcheng.net/oaa/.

Detecting and segmenting salient objects from natural scenes, often referred to as salient object detection, has attracted great interest in computer vision. While many models have been proposed and several applications have emerged, a deep understanding of achievements and issues remains lacking. We aim to provide a comprehensive review of recent progress in salient object detection and situate this field among other closely related areas such as generic scene segmentation, object proposal generation, and saliency for fixation prediction. Covering 228 publications, we survey i) roots, key concepts, and tasks, ii) core techniques and main modeling trends, and iii) datasets and evaluation metrics for salient object detection. We also discuss open problems such as evaluation metrics and dataset bias in model performance, and suggest future research directions.

In this paper, we propose a simple but effective framework for lane boundary detection, called SpinNet. Considering that cars or pedestrians often occlude lane boundaries and that the local features of lane boundaries are not distinctive, therefore, analyzing and collecting global context information is crucial for lane boundary detection. To this end, we design a novel spinning convolution layer and a brand-new lane parameterization branch in our network to detect lane boundaries from a global perspective. To extract features in narrow strip-shaped fields, we adopt strip-shaped convolutions with kernels which have 1 × n or n × 1 shape in the spinning convolution layer. To tackle the problem of that straight strip-shaped convolutions are only able to extract features in vertical or horizontal directions, we introduce the concept of feature map rotation to allow the convolutions to be applied in multiple directions so that more information can be collected concerning a whole lane boundary. Moreover, unlike most existing lane boundary detectors, which extract lane boundaries from segmentation masks, our lane boundary parameterization branch predicts a curve expression for the lane boundary for each pixel in the output feature map. And the network utilizes this information to predict the weights of the curve, to better form the final lane boundaries. Our framework is easy to implement and end-to-end trainable. Experiments show that our proposed SpinNet outperforms state-of-the-art methods.

Recent progress on salient object detection is substantial, benefiting mostly from the explosive development of Convolutional Neural Networks (CNNs). Semantic segmentation and salient object detection algorithms developed lately have been mostly based on Fully Convolutional Neural Networks (FCNs). There is still a large room for improvement over the generic FCN models that do not explicitly deal with the scale-space problem. The Holistically-Nested Edge Detector (HED) provides a skip-layer structure with deep supervision for edge and boundary detection, but the performance gain of HED on saliency detection is not obvious. In this paper, we propose a new salient object detection method by introducing short connections to the skip-layer structures within the HED architecture. Our framework takes full advantage of multi-level and multi-scale features extracted from FCNs, providing more advanced representations at each layer, a property that is critically needed to perform segment detection. Our method produces state-of-the-art results on 5 widely tested salient object detection benchmarks, with advantages in terms of efficiency (0.08 seconds per image), effectiveness, and simplicity over the existing algorithms. Beyond that, we conduct an exhaustive analysis of the role of training data on performance. We provide a training set for future research and fair comparisons.

In this paper, we reconsider the clustering problem for image over-segmentation from a new perspective. We propose a novel search algorithm named “active search” which explicitly considers neighboring continuity. Based on this search method, we design a back-and-forth traversal strategy and a "joint" assignment and update step to speed up the algorithm. Compared to earlier works, such as Simple Linear Iterative Clustering (SLIC) and its follow-ups, who use fixed search regions and perform the assignment and the update step separately, our novel scheme reduces the iteration number before convergence, as well as improves boundary sensitivity of the over-segmentation results. Extensive evaluations on the Berkeley segmentation benchmark verify that our method outperforms competing methods under various evaluation metrics. In particular, lowest time cost is reported among existing methods (approximately 30 fps for a 481321 image on a single CPU core). To facilitate the development of over-segmentation, the code will be publicly available.

Recently, adversarial erasing for weakly-supervised object attention has been deeply studied due to its capability in localizing integral object regions. However, such a strategy raises one key problem that attention regions will gradually expand to non-object regions as training iterations continue, which significantly decreases the quality of the produced attention maps. To tackle such an issue as well as promote the quality of object attention, we introduce a simple yet effective Self-Erasing Network (SeeNet) to prohibit attentions from spreading to unexpected background regions. In particular, SeeNet leverages two self-erasing strategies to encourage networks to use reliable object and background cues for learning to attention. In this way, integral object regions can be effectively highlighted without including much more background regions. To test the quality of the generated attention maps, we employ the mined object regions as heuristic cues for learning semantic segmentation models. Experiments on Pascal VOC well demonstrate the superiority of our SeeNet over other state-of-the-art methods.

In this paper, we reconsider the clustering problem for image over-segmentation from a new perspective. We propose a novel search algorithm called “active search” which explicitly considers neighbor continuity. Based on this search method, we design a back-and-forth traversal strategy and a joint assignment and update step to speed up the algorithm. Compared to earlier methods, such as simple linear iterative clustering (SLIC) and its variants, which use fixed search regions and perform the assignment and the update steps separately, our novel scheme reduces the number of iterations required for convergence, and also provides better boundaries in the over-segmentation results. Extensive evaluation using the Berkeley segmentation benchmark verifies that our method outperforms competing methods under various evaluation metrics. In particular, our method is fastest, achieving approximately 30 fps for a 481 × 321 image on a single CPU core. To facilitate further research, our code is made publicly available.

Recent progress on saliency detection is substantial, benefiting mostly from the explosive development of Convolutional Neural Networks (CNNs). Semantic segmentation and saliency detection algorithms developed lately have been mostly based on Fully Convolutional Neural Networks (FCNs). There is still a large room for improvement over the generic FCN models that do not explicitly deal with the scale-space problem. Holisitcally-Nested Edge Detector (HED) provides a skip-layer structure with deep supervision for edge and boundary detection, but the performance gain of HED on saliency detection is not obvious. In this paper, we propose a new saliency method by introducing short connections to the skip-layer structures within the HED architecture. Our framework provides rich multi-scale feature maps at each layer, a property that is critically needed to perform segment detection. Our method produces state-of-the-art results on 5 widely tested salient object detection benchmarks, with advantages in terms of efficiency (0.08 seconds per image), effectiveness, and simplicity over the existing algorithms.