Transformers and masked language modeling are quickly being adopted and explored in computer vision as vision transformers and masked image modeling (MIM). In this work, we argue that image token masking differs from token masking in text, due to the amount and correlation of tokens in an image. In particular, to generate a challenging pretext task for MIM, we advocate a shift from random masking to informed masking. We develop and exhibit this idea in the context of distillation-based MIM, where a teacher transformer encoder generates an attention map, which we use to guide masking for the student. We thus introduce a novel masking strategy, called attention-guided masking (AttMask), and we demonstrate its effectiveness over random masking for dense distillation-based MIM as well as plain distillation-based self-supervised learning on classification tokens. We confirm that AttMask accelerates the learning process and improves the performance on a variety of downstream tasks. We provide the implementation code at https://github.com/gkakogeorgiou/attmask.
论文检索
输入标题、作者或关键词，从 7,876 篇学术成果中精准定位
Image BERT pre-training with masked image modeling (MIM) becomes a popular practice to cope with self-supervised representation learning. A seminal work, BEiT, casts MIM as a classification task with a visual vocabulary, tokenizing the continuous visual signals into discrete vision tokens using a pre-learned dVAE. Despite a feasible solution, the improper discretization hinders further improvements of image pre-training. Since image discretization has no ground-truth answers, we believe that the masked patch should not be assigned with a unique token id even if a better “tokenizer” can be obtained. In this work, we introduce an improved BERT-style image pre-training method, namely mc-BEiT, which performs MIM proxy tasks towards eased and refined multi-choice training objectives. Specifically, the multi-choice supervision for the masked image patches is formed by the soft probability vectors of the discrete token ids, which are predicted by the off-the-shelf image “tokenizer” and further refined by high-level inter-patch perceptions resorting to the observation that similar patches should share their choices. Extensive experiments on classification, segmentation, and detection tasks demonstrate the superiority of our method, e.g., the pre-trained ViT-B achieves 84.1% top-1 fine-tuning accuracy on ImageNet-1K classification, 49.2% AP^b and 44.0% AP^m of object detection and instance segmentation on COCO, 50.8% mIOU on ADE20K semantic segmentation, outperforming the competitive counterparts. The code is available at https://github.com/lixiaotong97/mc-BEiT.
Weakly Supervised Object Localization through Inter-class Feature Similarity and Intra-Class Appearance Consistency
PDF ↗Weakly supervised object localization (WSOL) aims at detecting objects through only image-level labels. Class activation maps (CAMs) are the commonly used features for WSOL. However, existing CAM-based methods tend to excessively pursue discriminative features for object recognition and hence ignore the feature similarities among different categories, thereby leading to CAMs incomplete for object localization. In addition, CAMs are sensitive to background noise due to over-dependence on the holistic classification. In this paper, we propose a simple but effective WSOL model (named ISIC) through Inter-class feature Similarity and Intra-class appearance Consistency. In practice, our ISIC model first proposes the inter-class feature similarity (ICFS) loss against the original cross entropy loss. Such an ICFS loss sufficiently leverages the shared features together with the discriminative features between different categories, which significantly reduces the model over-fitting risk to background noise and brings more complete object masks. Besides, instead of CAMs, a non-negative matrix factorization mask module is applied to extract object masks from multiple intra-class images. Thanks to intra-class appearance consistency, the achieved pseudo masks are more complete and robust. As a result, extensive experiments confirm that our ISIC model achieves state-of-the-art on both CUB-200 and ImageNet-1K benchmarks i.e., 97.3% and 70.0% GT-Known localization accuracy, respectively.
The prime challenge in unsupervised domain adaptation (DA) is to mitigate the domain shift between the source and target domains. Prior DA works show that pretext tasks could be used to mitigate this domain shift by learning domain invariant representations. However, in practice, we find that most existing pretext tasks are ineffective against other established techniques. Thus, we theoretically analyze how and when a subsidiary pretext task could be leveraged to assist the goal task of a given DA problem and develop objective subsidiary task suitability criteria. Based on this criteria, we devise a novel process of sticker intervention and cast sticker classification as a supervised subsidiary DA problem concurrent to the goal task unsupervised DA. Our approach not only improves goal task adaptation performance, but also facilitates privacy-oriented source-free DA i.e. without concurrent source-target access. Experiments on the standard Office-31, Office-Home, DomainNet, and VisDA benchmarks demonstrate our superiority for both single-source and multi-source source-free DA. Our approach also complements existing non-source-free works, achieving leading performance.
With the development of generative-based self-supervised learning (SSL) approaches like BeiT and MAE, how to learn good representations by masking random patches of the input image and reconstructing the missing information has grown in concern. However, BeiT and PeCo need a “pre-pretraining” stage to produce discrete codebooks for masked patches representing. MAE does not require a pre-training codebook process, but setting pixels as reconstruction targets may introduce an optimization gap between pre-training and downstream tasks that good reconstruction quality may not always lead to the high descriptive capability for the model. Considering the above issues, in this paper, we propose a simple Self-distillated masked AutoEncoder network, namely SdAE. SdAE consists of a student branch using an encoder-decoder structure to reconstruct the missing information, and a teacher branch producing latent representation of masked tokens. We also analyze how to build good views for the teacher branch to produce latent representation from the perspective of information bottleneck. After that, we propose a multi-fold masking strategy to provide multiple masked views with balanced information for boosting the performance, which can also reduce the computational complexity. Our approach generalizes well: with only 300 epochs pre-training, a vanilla ViT-Base model achieves an 84.1% fine-tuning accuracy on ImageNet-1k classification, 48.6 mIOU on ADE20K segmentation, and 48.9 mAP on COCO detection with only 300 epochs pre-training, which surpasses other methods by a considerable margin. Code is available at https://github.com/AbrahamYabo/SdAE.
Contrastive learning has shown great promise in the field of graph representation learning. By manually constructing positive/negative samples, most graph contrastive learning methods rely on the vector inner product based similarity metric to distinguish the samples for graph representation. However, the handcrafted sample construction (e.g., the perturbation on the nodes or edges of the graph) may not effectively capture the intrinsic local structures of the graph. Also, the vector inner product based similarity metric cannot fully exploit the local structures of the graph to characterize the graph difference well. To this end, in this paper, we propose a novel adaptive subgraph generation based contrastive learning framework for efficient and robust self-supervised graph representation learning, and the optimal transport distance is utilized as the similarity metric between the subgraphs. It aims to generate contrastive samples by capturing the intrinsic structures of the graph and distinguish the samples based on the features and structures of subgraphs simultaneously. Specifically, for each center node, by adaptively learning relation weights to the nodes of the corresponding neighborhood, we first develop a network to generate the interpolated subgraph. We then construct the positive and negative pairs of subgraphs from the same and different nodes, respectively. Finally, we employ two types of optimal transport distances (i.e., Wasserstein distance and Gromov-Wasserstein distance) to construct the structured contrastive loss. Extensive node classification experiments on benchmark datasets verify the effectiveness of our graph contrastive learning method.
A Simple Baseline for Open-Vocabulary Semantic Segmentation with Pre-trained Vision-Language Model
PDF ↗Recently, open-vocabulary image classification by vision language pre-training has demonstrated incredible achievements, that the model can classify arbitrary categories without seeing additional annotated images of that category. However, it is still unclear how to make the open-vocabulary recognition work well on broader vision problems. This paper targets open-vocabulary semantic segmentation by building it on an off-the-shelf pre-trained vision-language model, i.e., CLIP. However, semantic segmentation and the CLIP model perform on different visual granularity, that semantic segmentation processes on pixels while CLIP performs on images. To remedy the discrepancy in processing granularity, we refuse the use of the prevalent one-stage FCN based framework, and advocate a two-stage semantic segmentation framework, with the first stage extracting generalizable mask proposals and the second stage leveraging an image based CLIP model to perform open-vocabulary classification on the masked image crops which are generated in the first stage. Our experimental results show that this two-stage framework can achieve superior performance than FCN when trained only on COCO Stuff dataset and evaluated on other datasets without fine-tuning. Moreover, this simple framework also surpasses previous state-of-the-arts of zero-shot semantic segmentation by a large margin: +29.5 hIoU on the Pascal VOC 2012 dataset, and +8.9 hIoU on the COCO Stuff dataset. With its simplicity and strong performance, we hope this framework to serve as a baseline to facilitate future research. The code are made publicly available at https://github.com/MendelXu/zsseg.baseline.
The segmentation task has traditionally been formulated as a complete-label pixel classification task to predict a class for each pixel from a fixed number of predefined semantic categories shared by all images or videos. Yet, following this formulation, standard architectures will inevitably encounter various challenges under more realistic settings where the scope of categories scales up (e.g., beyond the level of 1k). On the other hand, in a typical image or video, only a few categories, i.e., a small subset of the complete label are present. Motivated by this intuition, in this paper, we propose to decompose segmentation into two sub-problems: (i) image-level or video-level multi-label classification and (ii) pixel-level rank-adaptive selected-label classification. Given an input image or video, our framework first conducts multi-label classification over the complete label, then sorts the complete label and selects a small subset according to their class confidence scores. We then use a rank-adaptive pixel classifier to perform the pixel-wise classification over only the selected labels, which uses a set of rank-oriented learnable temperature parameters to adjust the pixel classifications scores. Our approach is conceptually general and can be used to improve various existing segmentation frameworks by simply using a lightweight multi-label classification head and rank-adaptive pixel classifier. We demonstrate the effectiveness of our framework with competitive experimental results across four tasks, including image semantic segmentation, image panoptic segmentation, video instance segmentation, and video semantic segmentation. Especially, with our RankSeg, Mask2Former gains +0.8%/+0.7%/+0.7% on ADE20K panoptic segmentation/YouTubeVIS 2019 video instance segmentation/VSPW video semantic segmentation benchmarks respectively.
Laplacian Mesh Transformer: Dual Attention and Topology Aware Network for 3D Mesh Classification and Segmentation
PDF ↗Deep learning-based approaches for shape understanding and processing tasks have attracted considerable attention. Despite the great progress that has been made, the existing approaches fail to efficiently capture sophisticated structure information and critical part features simultaneously, limiting their capability of providing discriminative deep shape features. To address the above issue, we proposed a novel deep learning framework, Laplacian Mesh Transformer, to extract the critical structure and geometry features. We introduce a dual attention mechanism, where the $1^{\rm st}$ level self-attention mechanism is used to capture the structure and critical partial geometric information on the entire mesh, and the $2^{\rm nd}$ level is to learn the importance of structure and geometric information. More particularly, Laplacian spectral decomposition is adopted as our basic structure representation given its ability to describe shape topology. Our approach builds a hierarchical structure to process shape features from fine to coarse using the dual attention mechanism, which is stable under the isometric transformations. It enables an effective feature extraction that can tackle 3D meshes with complex structure and geometry efficiently in various shape analysis tasks, such as shape segmentation and classification. Extensive experiments on the standard benchmarks show that our method outperforms state-of-the-art methods.
Traditional semantic segmentation requires a large labeled image dataset and can only be predicted within predefined classes. Solving this problem of few-shot segmentation, which requires only a handful of annotations for the new target class, is important. However, with few-shot segmentation, the target class data distribution in the feature space is sparse and has low coverage because of the slight variations in the sample data. Setting the classification boundary that properly separates the target class from other classes is an impossible task. In particular, it is difficult to classify classes that are similar to the target class near the boundary. This study proposes the Interclass Prototype Relation Network (IPRNet), which improves the separation performance by reducing the similarity between other classes. We conducted extensive experiments with Pascal-5i and COCO-20i and showed that IPRNet provides the best segmentation performance compared with previous research.
Adversarial Erasing Framework via Triplet with Gated Pyramid Pooling Layer for Weakly Supervised Semantic Segmentation
PDF ↗Weakly supervised semantic segmentation (WSSS) has employed Class Activation Maps (CAMs) to localize the objects. However, the CAMs typically do not fit along the object boundaries and highlight only the most-discriminative regions. To resolve the problems, we propose a Gated Pyramid Pooling (GPP) layer which is a substitute for a Global Average Pooling (GAP) layer, and an Adversarial Erasing Framework via Triplet (AEFT). In the GPP layer, a feature pyramid is obtained by pooling the CAMs at multiple spatial resolutions, and then be aggregated into an attention for class prediction by gated convolution. With the process, CAMs are trained not only to capture the global context but also to preserve fine-details from the image. Meanwhile, the AEFT targets an over-expansion, a chronic problem of Adversarial Erasing (AE). Although AE methods expand CAMs by erasing the discriminative regions, they usually suffer from the over-expansion due to an absence of guidelines on when to stop erasing. We experimentally verify that the over-expansion is due to rigid classification, and metric learning can be a flexible remedy for it. AEFT is devised to learn the concept of erasing with the triplet loss between the input image, erased image, and negatively sampled image. With the GPP and AEFT, we achieve new state-of-the-art both on the PASCAL VOC 2012 val/test and MS-COCO 2014 val set by 70.9%/71.7% and 44.8% in mIoU, respectively.
SegPGD: An Effective and Efficient Adversarial Attack for Evaluating and Boosting Segmentation Robustness
PDF ↗Deep neural network-based image classifications are vulnerable to adversarial perturbations. The image classifications can be easily fooled by adding artificial small and imperceptible perturbations to input images. As one of the most effective defense strategies, adversarial training was proposed to address the vulnerability of classification models, where the adversarial examples are created and injected into training data during training. The attack and defense of classification models have been intensively studied in past years. Semantic segmentation, as an extension of classifications, has also received great attention recently. Recent work shows a large number of attack iterations are required to create effective adversarial examples to fool segmentation models. The observation makes both robustness evaluation and adversarial training on segmentation models challenging. In this work, we propose an effective and efficient segmentation attack method, dubbed SegPGD. Besides, we provide a convergence analysis to show the proposed SegPGD can create more effective adversarial examples than PGD under the same number of attack iterations. Furthermore, we propose to apply our SegPGD as the underlying attack method for segmentation adversarial training. Since SegPGD can create more effective adversarial examples, the adversarial training with our SegPGD can boost the robustness of segmentation models. Our proposals are also verified with experiments on popular Segmentation model architectures and standard segmentation datasets.
For Weakly-Supervised Semantic Segmentation (WSSS) with image-level annotation, mostly relies on the classification network to generate initial segmentation pseudo-labels. However, the optimization target of classification networks usually neglects the discrimination between different pixels, like insignificant foreground and background regions. In this paper, we propose an adaptive Spatial Binary Cross-Entropy (Spatial-BCE) Loss for WSSS, which aims to enhance the discrimination between pixels. In Spatial-BCE Loss, we calculate the loss independently for each pixel and heuristically assign the optimization directions for foreground and background pixels separately. An auxiliary self-supervised task is also proposed to guarantee the Spatial-BCE Loss working as envisaged. Meanwhile, to enhance the network’s generalization for different data distributions, we design an alternate training strategy to adaptively generate thresholds to divide the foreground and background. Benefiting from high-quality initial pseudo-labels by Spatial-BCE Loss, our method also reduce the reliance on post-processing, thereby simplifying the pipeline of WSSS. Our method is validated on the PASCAL VOC 2012 and COCO2014 datasets and achieves the new state-of-the-art.
Egocentric videos offer fine-grain information for high-fidelity modeling of human behaviors. Hands and interacting objects are one crucial aspect of understanding viewer’s behaviors and intentions. We provide a labeled dataset consisting of 11,235 egocentric images with per-pixel segmentation labels of hands and the interacting objects in diverse daily activities. Our dataset is the first to label detailed interacting hand-object contact boundaries. We introduce a context-aware compositional data augmentation technique to adapt to out-of-the-distribution YouTube egocentric video. We show that our robust hand-object segmentation model and dataset can serve as a foundation tool to boost or enable several downstream vision applications, such as: Hand state classification, video activity recognition, 3D mesh reconstruction of hand-object interaction, and Seeing through the hand with video inpainting in egocentric videos. All of our data and code will be released to the public.
Parts represent a basic unit of geometric and semantic similarity across different objects. We argue that part knowledge should be composable beyond the observed object classes. Towards this, we present 3D Compositional Zero-shot Learning as a problem of part generalization from seen to unseen object classes for semantic segmentation. We provide a structured study through benchmarking the task with the proposed Compositional-PartNet dataset. This dataset is created by processing the original PartNet to maximize part overlap across different objects. The existing point cloud part segmentation methods fail to generalize to unseen object classes in this setting. As a solution, we propose DeCompositional Consensus, which combines a part segmentation network with a part scoring network. The key intuition to our approach is that a segmentation mask over some parts should have a consensus with its part scores when each part is taken apart. The two networks reason over different part combinations defined in a per-object part prior to generate the most suitable segmentation mask. We demonstrate that our method allows compositional zero-shot segmentation and generalized zero-shot classification and establishes the state of the art on both tasks.
Contrastive Language-Image Pre-training (CLIP) has made a remarkable breakthrough in open-vocabulary zero-shot image recognition. Many recent studies leverage the pre-trained CLIP models for image-level classification and manipulation. In this paper, we wish examine the intrinsic potential of CLIP for pixel-level dense prediction, specifically in semantic segmentation. To this end, with minimal modification, we show that MaskCLIP yields compelling segmentation results on open concepts across various datasets in the absence of annotations and fine-tuning. By adding pseudo labeling and self-training, MaskCLIP+ surpasses SOTA transductive zero-shot semantic segmentation methods by large margins, e.g., mIoUs of unseen classes on PASCAL VOC/PASCAL Context/COCO Stuff are improved from 35.6/20.7/30.3 to 86.1/66.7/54.7. We also test the robustness of MaskCLIP under input corruption and evaluate its capability in discriminating fine-grained objects and novel concepts. Our finding suggests that MaskCLIP can serve as a new reliable source of supervision for dense prediction tasks to achieve annotation-free segmentation. Source code is available at https://github.com/chongzhou96/MaskCLIP.
Most existing scene text detectors focus on detecting characters or words that only capture partial text messages due to missing contextual information. For a better understanding of text in scenes, it is more desired to detect contextual text blocks (CTBs) which consist of one or multiple integral text units (e.g., characters, words, or phrases) in natural reading order and transmit certain complete text messages. This paper presents contextual text detection, a new setup that detects CTBs for better understanding of texts in scenes. We formulate the new setup by a dual detection task which first detects integral text units and then groups them into a CTB. To this end, we design a novel scene text clustering technique that treats integral text units as tokens and groups them (belonging to the same CTB) into an ordered token sequence. In addition, we create two datasets SCUT-CTW-Context and ReCTS-Context to facilitate future research, where each CTB is well annotated by an ordered sequence of integral text units. Further, we introduce three metrics that measure contextual text detection in local accuracy, continuity, and global accuracy. Extensive experiments show that our method accurately detects CTBs which effectively facilitates downstream tasks such as text classification and translation. The project is available at \url{https://sg-vilab.github.io/publication/xue2022contextual/}.
Artistic text recognition is an extremely challenging task with a wide range of applications. However, current scene text recognition methods mainly focus on irregular text while have not explored artistic text specifically. The challenges of artistic text recognition include the various appearance with special-designed fonts and effects, the complex connections and overlaps between characters, and the severe interference from background patterns. To alleviate these problems, we propose to recognize the artistic text at three levels. Firstly, corner points are applied to guide the extraction of local features inside characters, considering the robustness of corner structures to appearance and shape. In this way, the discreteness of the corner points cuts off the connection between characters, and the sparsity of them improves the robustness for background interference. Secondly, we design a character contrastive loss to model the character-level feature, improving the feature representation for character classification. Thirdly, we utilize Transformer to learn the global feature on image-level and model the global relationship of the corner points, with the assistance of a corner-query cross-attention mechanism. Besides, we provide an artistic text dataset to benchmark the performance. Experimental results verify the significant superiority of our proposed method on artistic text recognition and also achieve state-of-the-art performance on several blurred and perspective datasets.
Text manipulation technologies cause serious worries in recent years, however, corresponding tampering detection methods have not been well explored. In this paper, we introduce a new task, named Tampered Scene Text Detection (TSTD), to localize text instances and recognize the texture authenticity in an end-to-end manner. Different from the general scene text detection (STD) task, TSTD further introduces the fine-grained classification, i.e. the tampered and real-world texts share a semantic space (text position and geometric structure) but have different local textures. To this end, we propose a simple yet effective modification strategy to migrate existing STD methods to TSTD task, keeping the semantic invariance while explicitly guiding the class-specific texture feature learning. Furthermore, we discuss the potential of frequency information for distinguishing feature learning, and propose a parallel-branch feature extractor to enhance the feature representation capability. To evaluate the effectiveness of our method, a new TSTD dataset (Tampered-IC13) is proposed and released at https://github.com/wangyuxin87/Tampered-IC13.
Panoptic Part Segmentation (PPS) aims to unify panoptic segmentation and part segmentation into one task. Previous work mainly utilizes separated approaches to handle thing, stuff, and part predictions individually without performing any shared computation and task association. In this work, we aim to unify these tasks at the architectural level, designing the first end-to-end unified method named Panoptic-PartFormer. In particular, motivated by the recent progress in Vision Transformer, we model things, stuff, and part as object queries and directly learn to optimize the all three predictions as unified mask prediction and classification problem. We design a decoupled decoder to generate part feature and thing/stuff feature respectively. Then we propose to utilize all the queries and corresponding features to perform reasoning jointly and iteratively. The final mask can be obtained via inner product between queries and the corresponding features. The extensive ablation studies and analysis prove the effectiveness of our framework. Our Panoptic-PartFormer achieves the new state-of-the-art results on both Cityscapes PPS and Pascal Context PPS datasets with around 70\% GFlops and 50\% parameters decrease. Given its effectiveness and conceptual simplicity, we hope the Panoptic-PartFormer can serve as a strong baseline and aid future research in PPS. Our code and models will be available at \url{https://github.com/lxtGH/Panoptic-PartFormer}.