The development of autoregressive modeling (AM) in computer vision lags behind natural language processing (NLP) in self-supervised pre-training. This is mainly caused by the challenge that images are not sequential signals and lack a natural order when applying autoregressive modeling. In this study, inspired by human beings’ way of grasping an image, i.e., focusing on the main object first, we present a semantic-aware autoregressive image modeling (SemAIM) method to tackle this challenge. The key insight of SemAIM is to autoregressively model images from the semantic patches to the less semantic patches. To this end, we first calculate a semantic-aware permutation of patches according to their feature similarities and then perform the autoregression procedure based on the permutation. In addition, considering that the raw pixels of patches are low-level signals and are not ideal prediction targets for learning high-level semantic representation, we also explore utilizing the patch features as the prediction targets. Extensive experiments are conducted on a broad range of downstream tasks, including image classification, object detection, and instance/semantic segmentation, to evaluate the performance of SemAIM. The results demonstrate SemAIM achieves state-of-the-art performance compared with other self-supervised methods. Specifically, with ViT-B, SemAIM achieves 84.1% top-1 accuracy for fine-tuning on ImageNet, 51.3% AP and 45.4% AP for object detection and instance segmentation on COCO, which outperforms the vanilla MAE by 0.5%, 1.0%, and 0.5%, respectively. Code is available at https://github.com/skyoux/SemAIM.
论文检索
输入标题、作者或关键词，从 5,999 篇学术成果中精准定位
The performance of existing unsupervised video object segmentation methods typically suffers from severe performance degradation on test videos when tested in out-of-distribution scenarios. The primary reason is that the test data in real- world may not follow the independent and identically distribution (i.i.d.) assumption, leading to domain shift. In this paper, we propose a generalizable fourier augmentation method during training to improve the generalization ability of the model. To achieve this, we perform Fast Fourier Transform (FFT) over the intermediate spatial domain features in each layer to yield corresponding frequency representations, including amplitude components (encoding scene-aware styles such as texture, color, contrast of the scene) and phase components (encoding rich semantics). We produce a variety of style features via Gaussian sampling to augment the training data, thereby improving the generalization capability of the model. To further improve the cross-domain generalization performance of the model, we design a phase feature update strategy via exponential moving average using phase features from past frames in an online update manner, which could help the model to learn cross-domain-invariant features. Extensive experiments show that our proposed method achieves the state-of-the-art performance on popular benchmarks.
In this work, we focus on label efficient learning for video action detection. We develop a novel semi-supervised active learning approach which utilizes both labeled as well as un- labeled data along with informative sample selection for ac- tion detection. Video action detection requires spatio-temporal localization along with classification, which poses several challenges for both active learning (informative sample se- lection) as well as semi-supervised learning (pseudo label generation). First, we propose NoiseAug, a simple augmenta- tion strategy which effectively selects informative samples for video action detection. Next, we propose fft-attention, a novel technique based on high-pass filtering which enables effective utilization of pseudo label for SSL in video action detection by emphasizing on relevant activity region within a video. We evaluate the proposed approach on three different bench- mark datasets, UCF-101-24, JHMDB-21, and Youtube-VOS. First, we demonstrate its effectiveness on video action detec- tion where the proposed approach outperforms prior works in semi-supervised and weakly-supervised learning along with several baseline approaches in both UCF101-24 and JHMDB- 21. Next, we also show its effectiveness on Youtube-VOS for video object segmentation demonstrating its generalization capability for other dense prediction tasks in videos.
Recent advances in deep learning have greatly improved the segmentation of mitochondria from Electron Microscopy (EM) images. However, suffering from variations in mitochondrial morphology, imaging conditions, and image noise, existing methods still exhibit high uncertainty in their predictions. Moreover, in view of our findings, predictions with high levels of uncertainty are often accompanied by inaccuracies such as ambiguous boundaries and amount of false positive segments. To deal with the above problems, we propose a novel approach for mitochondria segmentation in 3D EM images that leverages evidential uncertainty estimation, which for the first time integrates evidential uncertainty to enhance the performance of segmentation. To be more specific, our proposed method not only provides accurate segmentation results, but also estimates associated uncertainty. Then, the estimated uncertainty is used to help improve the segmentation performance by an uncertainty rectification module, which leverages uncertainty maps and multi-scale information to refine the segmentation. Extensive experiments conducted on four challenging benchmarks demonstrate the superiority of our proposed method over existing approaches.
Few-shot semantic segmentation (FSS) aims to segment unseen objects in a query image using a few pixel-wise annotated support images, thus expanding the capabilities of semantic segmentation. The main challenge lies in extracting sufficient information from the limited support images to guide the segmentation process. Conventional methods typically address this problem by generating single or multiple prototypes from the support images and calculating their cosine similarity to the query image. However, these methods often fail to capture meaningful information for modeling the de facto joint distribution of pixel and category. Consequently, they result in incomplete segmentation of foreground objects and mis-segmentation of the complex background. To overcome this issue, we propose the Cross Gaussian Mixture Generative Model (CGMGM), a novel Gaussian Mixture Models~(GMMs)-based FSS method, which establishes the joint distribution of pixel and category in both the support and query images. Specifically, our method initially matches the feature representations of the query image with those of the support images to generate and refine an initial segmentation mask. It then employs GMMs to accurately model the joint distribution of foreground and background using the support masks and the initial segmentation mask. Subsequently, a parametric decoder utilizes the posterior probability of pixels in the query image, by applying the Bayesian theorem, to the joint distribution, to generate the final segmentation mask. Experimental results on PASCAL-5i and COCO-20i datasets demonstrate our CGMGM's effectiveness and superior performance compared to the state-of-the-art methods.
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.
The dot product self-attention (DPSA) is a fundamental component of transformers. However, scaling them to long sequences, like documents or high-resolution images, becomes prohibitively expensive due to the quadratic time and memory complexities arising from the softmax operation. Kernel methods are employed to simplify computations by approximating softmax but often lead to performance drops compared to softmax attention. We propose SeTformer, a novel transformer where DPSA is purely replaced by Self-optimal Transport (SeT) for achieving better performance and computational efficiency. SeT is based on two essential softmax properties: maintaining a non-negative attention matrix and using a nonlinear reweighting mechanism to emphasize important tokens in input sequences. By introducing a kernel cost function for optimal transport, SeTformer effectively satisfies these properties. In particular, with small and base-sized models, SeTformer achieves impressive top-1 accuracies of 84.7% and 86.2% on ImageNet-1K. In object detection, SeTformer-base outperforms the FocalNet counterpart by +2.2 mAP, using 38% fewer parameters and 29% fewer FLOPs. In semantic segmentation, our base-size model surpasses NAT by +3.5 mIoU with 33% fewer parameters. SeTformer also achieves state-of-the-art results in language modeling on the GLUE benchmark. These findings highlight SeTformer applicability for vision and language tasks.
X-RefSeg3D: Enhancing Referring 3D Instance Segmentation via Structured Cross-Modal Graph Neural Networks
PDF ↗Referring 3D instance segmentation is a challenging task aimed at accurately segmenting a target instance within a 3D scene based on a given referring expression. However, previous methods have overlooked the distinct roles played by different words in referring expressions. Additionally, they have failed to incorporate the positional relationship within referring expressions with the spatial correlations in 3D scenes. To alleviate these issues, we present a novel model called X-RefSeg3D, which constructs a cross-modal graph for the input 3D scene and unites textual and spatial relationships for reasoning via graph neural networks. Our approach begins by capturing object-specific text features, which are then fused with the instance features to construct a comprehensive cross-modal scene graph. Subsequently, we integrate the obtained cross-modal features into graph neural networks, leveraging the K-nearest algorithm to derive explicit instructions from expressions and factual relationships in scenes. This enables the effective capture of higher-order relationships among instances, thereby enhancing feature fusion and facilitating reasoning. Finally, the refined feature undergoes a matching module to compute the ultimate matching score. Experimental results on ScanRefer demonstrate the effectiveness of our method, surpassing previous approaches by a substantial margin of +3.67% in terms of mIOU.
Segment Anything Model (SAM) has received remarkable attention as it offers a powerful and versatile solution for object segmentation in images. However, fine-tuning SAM for downstream segmentation tasks under different scenarios remains a challenge, as the varied characteristics of different scenarios naturally requires diverse model parameter spaces. Most existing fine-tuning methods attempt to bridge the gaps among different scenarios by introducing a set of new parameters to modify SAM's original parameter space. Unlike these works, in this paper, we propose fine-tuning SAM efficiently by parameter space reconstruction (SAM-PARSER), which introduce nearly zero trainable parameters during fine-tuning. In SAM-PARSER, we assume that SAM's original parameter space is relatively complete, so that its bases are able to reconstruct the parameter space of a new scenario. We obtain the bases by matrix decomposition, and fine-tuning the coefficients to reconstruct the parameter space tailored to the new scenario by an optimal linear combination of the bases. Experimental results show that SAM-PARSER exhibits superior segmentation performance across various scenarios, while reducing the number of trainable parameters by approximately 290 times compared with current parameter-efficient fine-tuning methods.
ViTEraser: Harnessing the Power of Vision Transformers for Scene Text Removal with SegMIM Pretraining
PDF ↗Scene text removal (STR) aims at replacing text strokes in natural scenes with visually coherent backgrounds. Recent STR approaches rely on iterative refinements or explicit text masks, resulting in high complexity and sensitivity to the accuracy of text localization. Moreover, most existing STR methods adopt convolutional architectures while the potential of vision Transformers (ViTs) remains largely unexplored. In this paper, we propose a simple-yet-effective ViT-based text eraser, dubbed ViTEraser. Following a concise encoder-decoder framework, ViTEraser can easily incorporate various ViTs to enhance long-range modeling. Specifically, the encoder hierarchically maps the input image into the hidden space through ViT blocks and patch embedding layers, while the decoder gradually upsamples the hidden features to the text-erased image with ViT blocks and patch splitting layers. As ViTEraser implicitly integrates text localization and inpainting, we propose a novel end-to-end pretraining method, termed SegMIM, which focuses the encoder and decoder on the text box segmentation and masked image modeling tasks, respectively. Experimental results demonstrate that ViTEraser with SegMIM achieves state-of-the-art performance on STR by a substantial margin and exhibits strong generalization ability when extended to other tasks, e.g., tampered scene text detection. Furthermore, we comprehensively explore the architecture, pretraining, and scalability of the ViT-based encoder-decoder for STR, which provides deep insights into the application of ViT to the STR field. Code is available at https://github.com/shannanyinxiang/ViTEraser.
Few-shot segmentation aims to accurately segment novel target objects within query images using only a limited number of annotated support images. The recent works exploit support background as well as its foreground to precisely compute the dense correlations between query and support. However, they overlook the characteristics of the background that generally contains various types of objects. In this paper, we highlight this characteristic of background which can bring problematic cases as follows: (1) when the query and support backgrounds are dissimilar and (2) when objects in the support background are similar to the target object in the query. Without any consideration of the above cases, adopting the entire support background leads to a misprediction of the query foreground as background. To address this issue, we propose Task-disruptive Background Suppression(TBS), a module to suppress those disruptive support background features based on two spatial-wise scores: query-relevant and target-relevant scores. The former aims to mitigate the impact of unshared features solely existing in the support background, while the latter aims to reduce the influence of target-similar support background features. Based on these two scores, we define a query background relevant score that captures the similarity between the backgrounds of the query and the support, and utilize it to scale support background features to adaptively restrict the impact of disruptive support backgrounds. Our proposed method achieves state-of-the-art performance on standard few-shot segmentation benchmarks. Our official code is available at github.com/SuhoPark0706/TBSNet.
RadarMOSEVE: A Spatial-Temporal Transformer Network for Radar-Only Moving Object Segmentation and Ego-Velocity Estimation
PDF ↗Moving object segmentation (MOS) and Ego velocity estimation (EVE) are vital capabilities for mobile systems to achieve full autonomy. Several approaches have attempted to achieve MOSEVE using a LiDAR sensor. However, LiDAR sensors are typically expensive and susceptible to adverse weather conditions. Instead, millimeter-wave radar (MWR) has gained popularity in robotics and autonomous driving for real applications due to its cost-effectiveness and resilience to bad weather. Nonetheless, publicly available MOSEVE datasets and approaches using radar data are limited. Some existing methods adopt point convolutional networks from LiDAR-based approaches, ignoring the specific artifacts and the valuable radial velocity information of radar measurements, leading to suboptimal performance. In this paper, we propose a novel transformer network that effectively addresses the sparsity and noise issues and leverages the radial velocity measurements of radar points using our devised radar self- and cross-attention mechanisms. Based on that, our method achieves accurate EVE of the robot and performs MOS using only radar data simultaneously. To thoroughly evaluate the MOSEVE performance of our method, we annotated the radar points in the public View-of-Delft (VoD) dataset and additionally constructed a new radar dataset in various environments. The experimental results demonstrate the superiority of our approach over existing state-of-the-art methods. The code is available at https://github.com/ORCAUboat/RadarMOSEVE.
Weak supervision has proven to be an effective strategy for reducing the burden of annotating semantic segmentation tasks in 3D space. However, unconstrained or heuristic weakly supervised annotation forms may lead to suboptimal label efficiency. To address this issue, we propose a novel label recommendation framework for weakly supervised point cloud semantic segmentation. Distinct from pre-training and active learning, the label recommendation framework consists of three stages: inductive bias learning, recommendations for points to be labeled, and point cloud semantic segmentation learning. In practice, we first introduce the point cloud upsampling task to induct inductive bias from structural information. During the recommendation stage, we present a cross-scene clustering strategy to generate centers of clustering as recommended points. Then we introduce a recommended point positions attention module LabelAttention to model the long-range dependency under sparse annotations. Additionally, we employ position encoding to enhance the spatial awareness of semantic features. Throughout the framework, the useful information obtained from inductive bias learning is propagated to subsequent semantic segmentation networks in the form of label positions. Experimental results demonstrate that our framework outperforms weakly supervised point cloud semantic segmentation methods and other methods for labeling efficiency on S3DIS and ScanNetV2, even at an extremely low label rate.
Interactive segmentation methods rely on user inputs to iteratively update the selection mask. A click specifying the object of interest is arguably the most simple and intuitive interaction type, and thereby the most common choice for interactive segmentation. However, user clicking patterns in the interactive segmentation context remain unexplored. Accordingly, interactive segmentation evaluation strategies rely more on intuition and common sense rather than empirical studies (e.g., assuming that users tend to click in the center of the area with the largest error). In this work, we conduct a real-user study to investigate real user clicking patterns. This study reveals that the intuitive assumption made in the common evaluation strategy may not hold. As a result, interactive segmentation models may show high scores in the standard benchmarks, but it does not imply that they would perform well in a real world scenario. To assess the applicability of interactive segmentation methods, we propose a novel evaluation strategy providing a more comprehensive analysis of a model's performance. To this end, we propose a methodology for finding extreme user inputs by a direct optimization in a white-box adversarial attack on the interactive segmentation model. Based on the performance with such adversarial user inputs, we assess the robustness of interactive segmentation models w.r.t click positions. Besides, we introduce a novel benchmark for measuring the robustness of interactive segmentation, and report the results of an extensive evaluation of dozens of models.
NaMa: Neighbor-Aware Multi-Modal Adaptive Learning for Prostate Tumor Segmentation on Anisotropic MR Images
PDF ↗Accurate segmentation of prostate tumors from multi-modal magnetic resonance (MR) images is crucial for diagnosis and treatment of prostate cancer. However, the robustness of existing segmentation methods is limited, mainly because these methods 1) fail to adaptively assess subject-specific information of each MR modality for accurate tumor delineation, and 2) lack effective utilization of inter-slice information across thick slices in MR images to segment tumor as a whole 3D volume. In this work, we propose a two-stage neighbor-aware multi-modal adaptive learning network (NaMa) for accurate prostate tumor segmentation from multi-modal anisotropic MR images. In particular, in the first stage, we apply subject-specific multi-modal fusion in each slice by developing a novel modality-informativeness adaptive learning (MIAL) module for selecting and adaptively fusing informative representation of each modality based on inter-modality correlations. In the second stage, we exploit inter-slice feature correlations to derive volumetric tumor segmentation. Specifically, we first use a Unet variant with sequence layers to coarsely capture slice relationship at a global scale, and further generate an activation map for each slice. Then, we introduce an activation mapping guidance (AMG) module to refine slice-wise representation (via information from adjacent slices) for consistent tumor segmentation across neighboring slices. Besides, during the network training, we further apply a random mask strategy to each MR modality to improve feature representation efficiency. Experiments on both in-house and public (PICAI) multi-modal prostate tumor datasets show that our proposed NaMa performs better than state-of-the-art methods.
Pay Attention to Target: Relation-Aware Temporal Consistency for Domain Adaptive Video Semantic Segmentation
PDF ↗Video semantic segmentation has achieved conspicuous achievements attributed to the development of deep learning, but suffers from labor-intensive annotated training data gathering. To alleviate the data-hunger issue, domain adaptation approaches are developed in the hope of adapting the model trained on the labeled synthetic videos to the real videos in the absence of annotations. By analyzing the dominant paradigm consistency regularization in the domain adaptation task, we find that the bottlenecks exist in previous methods from the perspective of pseudo-labels. To take full advantage of the information contained in the pseudo-labels and empower more effective supervision signals, we propose a coherent PAT network including a target domain focalizer and relation-aware temporal consistency. The proposed PAT network enjoys several merits. First, the target domain focalizer is responsible for paying attention to the target domain, and increasing the accessibility of pseudo-labels in consistency training. Second, the relation-aware temporal consistency aims at modeling the inter-class consistent relationship across frames to equip the model with effective supervision signals. Extensive experimental results on two challenging benchmarks demonstrate that our method performs favorably against state-of-the-art domain adaptive video semantic segmentation methods.
Video-language pre-training models have recently achieved remarkable results on various multi-modal downstream tasks. However, most of these models rely on contrastive learning or masking modeling to align global features across modalities, neglecting the local associations between video frames and text tokens. This limits the model’s ability to perform fine-grained matching and generalization, especially for tasks that selecting segments in long videos based on query texts. To address this issue, we propose a novel stitching and matching pre-text task for video-language pre-training that encourages fine-grained interactions between modalities. Our task involves stitching video frames or sentences into longer sequences and predicting the positions of cross-model queries in the stitched sequences. The individual frame and sentence representations are thus aligned via the stitching and matching strategy, encouraging the fine-grained interactions between videos and texts. in the stitched sequences for the cross-modal query. We conduct extensive experiments on various benchmarks covering text-to-video retrieval, video question answering, video captioning, and moment retrieval. Our results demonstrate that the proposed method significantly improves the generalization capacity of the video-text pre-training models.
Federated learning collaboratively trains machine learning models among different clients while keeping data privacy and has become the mainstream for breaking data silos. However, the non-independently and identically distribution (i.e., Non-IID) characteristic of different image domains among different clients reduces the benefits of federated learning and has become a bottleneck problem restricting the accuracy and generalization of federated models. In this work, we propose a novel federated image segmentation method based on style transfer, FedST, by using a denoising diffusion probabilistic model to achieve feature disentanglement and image synthesis of cross-domain image data between multiple clients. Thus it can share style features among clients while protecting structure features of image data, which effectively alleviates the influence of the Non-IID phenomenon. Experiments prove that our method achieves superior segmentation performance compared to state-of-art methods among four different Non-IID datasets in objective and subjective assessment. The code is available at https://github.com/YoferChen/FedST.
Automatic mitochondrial segmentation enjoys great popularity with the development of deep learning. However, the coarse prediction raised by the presence of regular 3D grids in previous methods regardless of 3D CNN or the vision transformers suggest a possibly sub-optimal feature arrangement. To mitigate this limitation, we attempt to interpret the 3D EM image stacks as a set of interrelated 3D fragments for a better solution. However, it is non-trivial to model the 3D fragments without introducing excessive computational overhead. In this paper, we design a coherent fragment vision transformer (FragViT) combined with affinity learning to manipulate features on 3D fragments yet explore mutual relationships to model fragment-wise context, enjoying locality prior without sacrificing global reception. The proposed FragViT includes a fragment encoder and a hierarchical fragment aggregation module. The fragment encoder is equipped with affinity heads to transform the tokens into fragments with homogeneous semantics, and the multi-layer self-attention is used to explicitly learn inter-fragment relations with long-range dependencies. The hierarchical fragment aggregation module is responsible for hierarchically aggregating fragment-wise prediction back to the final voxel-wise prediction in a progressive manner. Extensive experimental results on the challenging MitoEM, Lucchi, and AC3/AC4 benchmarks demonstrate the effectiveness of the proposed method.
Video Moment Retrieval (VMR) aims to retrieve temporal segments in untrimmed videos corresponding to a given language query by constructing cross-modal alignment strategies. However, these existing strategies are often sub-optimal since they ignore the modality imbalance problem, i.e., the semantic richness inherent in videos far exceeds that of a given limited-length sentence. Therefore, in pursuit of better alignment, a natural idea is enhancing the video modality to filter out query-irrelevant semantics, and enhancing the text modality to capture more segment-relevant knowledge. In this paper, we introduce Modal-Enhanced Semantic Modeling (MESM), a novel framework for more balanced alignment through enhancing features at two levels. First, we enhance the video modality at the frame-word level through word reconstruction. This strategy emphasizes the portions associated with query words in frame-level features while suppressing irrelevant parts. Therefore, the enhanced video contains less redundant semantics and is more balanced with the textual modality. Second, we enhance the textual modality at the segment-sentence level by learning complementary knowledge from context sentences and ground-truth segments. With the knowledge added to the query, the textual modality thus maintains more meaningful semantics and is more balanced with the video modality. By implementing two levels of MESM, the semantic information from both modalities is more balanced to align, thereby bridging the modality gap. Experiments on three widely used benchmarks, including the out-of-distribution settings, show that the proposed framework achieves a new start-of-the-art performance with notable generalization ability (e.g., 4.42% and 7.69% average gains of R1@0.7 on Charades-STA and Charades-CG). The code will be available at https://github.com/lntzm/MESM.