2022-06-25 更新
Patch-level Representation Learning for Self-supervised Vision Transformers
Authors:Sukmin Yun, Hankook Lee, Jaehyung Kim, Jinwoo Shin
Recent self-supervised learning (SSL) methods have shown impressive results in learning visual representations from unlabeled images. This paper aims to improve their performance further by utilizing the architectural advantages of the underlying neural network, as the current state-of-the-art visual pretext tasks for SSL do not enjoy the benefit, i.e., they are architecture-agnostic. In particular, we focus on Vision Transformers (ViTs), which have gained much attention recently as a better architectural choice, often outperforming convolutional networks for various visual tasks. The unique characteristic of ViT is that it takes a sequence of disjoint patches from an image and processes patch-level representations internally. Inspired by this, we design a simple yet effective visual pretext task, coined SelfPatch, for learning better patch-level representations. To be specific, we enforce invariance against each patch and its neighbors, i.e., each patch treats similar neighboring patches as positive samples. Consequently, training ViTs with SelfPatch learns more semantically meaningful relations among patches (without using human-annotated labels), which can be beneficial, in particular, to downstream tasks of a dense prediction type. Despite its simplicity, we demonstrate that it can significantly improve the performance of existing SSL methods for various visual tasks, including object detection and semantic segmentation. Specifically, SelfPatch significantly improves the recent self-supervised ViT, DINO, by achieving +1.3 AP on COCO object detection, +1.2 AP on COCO instance segmentation, and +2.9 mIoU on ADE20K semantic segmentation.
PDF Accepted to CVPR 2022. Code is available at https://github.com/alinlab/SelfPatch
论文截图
SP-ViT: Learning 2D Spatial Priors for Vision Transformers
Authors:Yuxuan Zhou, Wangmeng Xiang, Chao Li, Biao Wang, Xihan Wei, Lei Zhang, Margret Keuper, Xiansheng Hua
Recently, transformers have shown great potential in image classification and established state-of-the-art results on the ImageNet benchmark. However, compared to CNNs, transformers converge slowly and are prone to overfitting in low-data regimes due to the lack of spatial inductive biases. Such spatial inductive biases can be especially beneficial since the 2D structure of an input image is not well preserved in transformers. In this work, we present Spatial Prior-enhanced Self-Attention (SP-SA), a novel variant of vanilla Self-Attention (SA) tailored for vision transformers. Spatial Priors (SPs) are our proposed family of inductive biases that highlight certain groups of spatial relations. Unlike convolutional inductive biases, which are forced to focus exclusively on hard-coded local regions, our proposed SPs are learned by the model itself and take a variety of spatial relations into account. Specifically, the attention score is calculated with emphasis on certain kinds of spatial relations at each head, and such learned spatial foci can be complementary to each other. Based on SP-SA we propose the SP-ViT family, which consistently outperforms other ViT models with similar GFlops or parameters. Our largest model SP-ViT-L achieves a record-breaking 86.3% Top-1 accuracy with a reduction in the number of parameters by almost 50% compared to previous state-of-the-art model (150M for SP-ViT-L vs 271M for CaiT-M-36) among all ImageNet-1K models trained on 224x224 and fine-tuned on 384x384 resolution w/o extra data.
PDF
论文截图
Self-supervised Vision Transformers for Joint SAR-optical Representation Learning
Authors:Yi Wang, Conrad M Albrecht, Xiao Xiang Zhu
Self-supervised learning (SSL) has attracted much interest in remote sensing and earth observation due to its ability to learn task-agnostic representations without human annotation. While most of the existing SSL works in remote sensing utilize ConvNet backbones and focus on a single modality, we explore the potential of vision transformers (ViTs) for joint SAR-optical representation learning. Based on DINO, a state-of-the-art SSL algorithm that distills knowledge from two augmented views of an input image, we combine SAR and optical imagery by concatenating all channels to a unified input. Subsequently, we randomly mask out channels of one modality as a data augmentation strategy. While training, the model gets fed optical-only, SAR-only, and SAR-optical image pairs learning both inner- and intra-modality representations. Experimental results employing the BigEarthNet-MM dataset demonstrate the benefits of both, the ViT backbones and the proposed multimodal SSL algorithm DINO-MM.
PDF 4 pages, 1 figure; IGARSS 2022
论文截图
DualCoOp: Fast Adaptation to Multi-Label Recognition with Limited Annotations
Authors:Ximeng Sun, Ping Hu, Kate Saenko
Solving multi-label recognition (MLR) for images in the low-label regime is a challenging task with many real-world applications. Recent work learns an alignment between textual and visual spaces to compensate for insufficient image labels, but loses accuracy because of the limited amount of available MLR annotations. In this work, we utilize the strong alignment of textual and visual features pretrained with millions of auxiliary image-text pairs and propose Dual Context Optimization (DualCoOp) as a unified framework for partial-label MLR and zero-shot MLR. DualCoOp encodes positive and negative contexts with class names as part of the linguistic input (i.e. prompts). Since DualCoOp only introduces a very light learnable overhead upon the pretrained vision-language framework, it can quickly adapt to multi-label recognition tasks that have limited annotations and even unseen classes. Experiments on standard multi-label recognition benchmarks across two challenging low-label settings demonstrate the advantages of our approach over state-of-the-art methods.
PDF
论文截图
Masked Frequency Modeling for Self-Supervised Visual Pre-Training
Authors:Jiahao Xie, Wei Li, Xiaohang Zhan, Ziwei Liu, Yew Soon Ong, Chen Change Loy
We present Masked Frequency Modeling (MFM), a unified frequency-domain-based approach for self-supervised pre-training of visual models. Instead of randomly inserting mask tokens to the input embeddings in the spatial domain, in this paper, we shift the perspective to the frequency domain. Specifically, MFM first masks out a portion of frequency components of the input image and then predicts the missing frequencies on the frequency spectrum. Our key insight is that predicting masked components in the frequency domain is more ideal to reveal underlying image patterns rather than predicting masked patches in the spatial domain, due to the heavy spatial redundancy. Our findings suggest that with the right configuration of mask-and-predict strategy, both the structural information within high-frequency components and the low-level statistics among low-frequency counterparts are useful in learning good representations. For the first time, MFM demonstrates that, for both ViT and CNN, a simple non-Siamese framework can learn meaningful representations even using none of the following: (i) extra data, (ii) extra model, (iii) mask token. Experimental results on ImageNet and several robustness benchmarks show the competitive performance and advanced robustness of MFM compared with recent masked image modeling approaches. Furthermore, we also comprehensively investigate the effectiveness of classical image restoration tasks for representation learning from a unified frequency perspective and reveal their intriguing relations with our MFM approach. Project page: https://www.mmlab-ntu.com/project/mfm/index.html.
PDF Project page: https://www.mmlab-ntu.com/project/mfm/index.html
论文截图
Scaling Autoregressive Models for Content-Rich Text-to-Image Generation
Authors:Jiahui Yu, Yuanzhong Xu, Jing Yu Koh, Thang Luong, Gunjan Baid, Zirui Wang, Vijay Vasudevan, Alexander Ku, Yinfei Yang, Burcu Karagol Ayan, Ben Hutchinson, Wei Han, Zarana Parekh, Xin Li, Han Zhang, Jason Baldridge, Yonghui Wu
We present the Pathways Autoregressive Text-to-Image (Parti) model, which generates high-fidelity photorealistic images and supports content-rich synthesis involving complex compositions and world knowledge. Parti treats text-to-image generation as a sequence-to-sequence modeling problem, akin to machine translation, with sequences of image tokens as the target outputs rather than text tokens in another language. This strategy can naturally tap into the rich body of prior work on large language models, which have seen continued advances in capabilities and performance through scaling data and model sizes. Our approach is simple: First, Parti uses a Transformer-based image tokenizer, ViT-VQGAN, to encode images as sequences of discrete tokens. Second, we achieve consistent quality improvements by scaling the encoder-decoder Transformer model up to 20B parameters, with a new state-of-the-art zero-shot FID score of 7.23 and finetuned FID score of 3.22 on MS-COCO. Our detailed analysis on Localized Narratives as well as PartiPrompts (P2), a new holistic benchmark of over 1600 English prompts, demonstrate the effectiveness of Parti across a wide variety of categories and difficulty aspects. We also explore and highlight limitations of our models in order to define and exemplify key areas of focus for further improvements. See https://parti.research.google/ for high-resolution images.
PDF Preprint
论文截图
Backdoor Attacks on Vision Transformers
Authors:Akshayvarun Subramanya, Aniruddha Saha, Soroush Abbasi Koohpayegani, Ajinkya Tejankar, Hamed Pirsiavash
Vision Transformers (ViT) have recently demonstrated exemplary performance on a variety of vision tasks and are being used as an alternative to CNNs. Their design is based on a self-attention mechanism that processes images as a sequence of patches, which is quite different compared to CNNs. Hence it is interesting to study if ViTs are vulnerable to backdoor attacks. Backdoor attacks happen when an attacker poisons a small part of the training data for malicious purposes. The model performance is good on clean test images, but the attacker can manipulate the decision of the model by showing the trigger at test time. To the best of our knowledge, we are the first to show that ViTs are vulnerable to backdoor attacks. We also find an intriguing difference between ViTs and CNNs - interpretation algorithms effectively highlight the trigger on test images for ViTs but not for CNNs. Based on this observation, we propose a test-time image blocking defense for ViTs which reduces the attack success rate by a large margin. Code is available here: https://github.com/UCDvision/backdoor_transformer.git
PDF
论文截图
SemMAE: Semantic-Guided Masking for Learning Masked Autoencoders
Authors:Gang Li, Heliang Zheng, Daqing Liu, Bing Su, Changwen Zheng
Recently, significant progress has been made in masked image modeling to catch up to masked language modeling. However, unlike words in NLP, the lack of semantic decomposition of images still makes masked autoencoding (MAE) different between vision and language. In this paper, we explore a potential visual analogue of words, i.e., semantic parts, and we integrate semantic information into the training process of MAE by proposing a Semantic-Guided Masking strategy. Compared to widely adopted random masking, our masking strategy can gradually guide the network to learn various information, i.e., from intra-part patterns to inter-part relations. In particular, we achieve this in two steps. 1) Semantic part learning: we design a self-supervised part learning method to obtain semantic parts by leveraging and refining the multi-head attention of a ViT-based encoder. 2) Semantic-guided MAE (SemMAE) training: we design a masking strategy that varies from masking a portion of patches in each part to masking a portion of (whole) parts in an image. Extensive experiments on various vision tasks show that SemMAE can learn better image representation by integrating semantic information. In particular, SemMAE achieves 84.5% fine-tuning accuracy on ImageNet-1k, which outperforms the vanilla MAE by 1.4%. In the semantic segmentation and fine-grained recognition tasks, SemMAE also brings significant improvements and yields the state-of-the-art performance.
PDF 12 pages, 3 figures
论文截图
Rectify ViT Shortcut Learning by Visual Saliency
Authors:Chong Ma, Lin Zhao, Yuzhong Chen, David Weizhong Liu, Xi Jiang, Tuo Zhang, Xintao Hu, Dinggang Shen, Dajiang Zhu, Tianming Liu
Shortcut learning is common but harmful to deep learning models, leading to degenerated feature representations and consequently jeopardizing the model’s generalizability and interpretability. However, shortcut learning in the widely used Vision Transformer framework is largely unknown. Meanwhile, introducing domain-specific knowledge is a major approach to rectifying the shortcuts, which are predominated by background related factors. For example, in the medical imaging field, eye-gaze data from radiologists is an effective human visual prior knowledge that has the great potential to guide the deep learning models to focus on meaningful foreground regions of interest. However, obtaining eye-gaze data is time-consuming, labor-intensive and sometimes even not practical. In this work, we propose a novel and effective saliency-guided vision transformer (SGT) model to rectify shortcut learning in ViT with the absence of eye-gaze data. Specifically, a computational visual saliency model is adopted to predict saliency maps for input image samples. Then, the saliency maps are used to distil the most informative image patches. In the proposed SGT, the self-attention among image patches focus only on the distilled informative ones. Considering this distill operation may lead to global information lost, we further introduce, in the last encoder layer, a residual connection that captures the self-attention across all the image patches. The experiment results on four independent public datasets show that our SGT framework can effectively learn and leverage human prior knowledge without eye gaze data and achieves much better performance than baselines. Meanwhile, it successfully rectifies the harmful shortcut learning and significantly improves the interpretability of the ViT model, demonstrating the promise of transferring human prior knowledge derived visual saliency in rectifying shortcut learning
PDF NeurIPS2022 Under Review
论文截图
Global Context Vision Transformers
Authors:Ali Hatamizadeh, Hongxu Yin, Jan Kautz, Pavlo Molchanov
We propose global context vision transformer (GC ViT), a novel architecture that enhances parameter and compute utilization. Our method leverages global context self-attention modules, joint with local self-attention, to effectively yet efficiently model both long and short-range spatial interactions, without the need for expensive operations such as computing attention masks or shifting local windows. In addition, we address the issue of lack of the inductive bias in ViTs via proposing to use a modified fused inverted residual blocks in our architecture. Our proposed GC ViT achieves state-of-the-art results across image classification, object detection and semantic segmentation tasks. On ImageNet-1K dataset for classification, the base, small and tiny variants of GC ViT with $28$M, $51$M and $90$M parameters achieve $\textbf{83.2\%}$, $\textbf{83.9\%}$ and $\textbf{84.4\%}$ Top-1 accuracy, respectively, surpassing comparably-sized prior art such as CNN-based ConvNeXt and ViT-based Swin Transformer by a large margin. Pre-trained GC ViT backbones in downstream tasks of object detection, instance segmentation, and semantic segmentation using MS COCO and ADE20K datasets outperform prior work consistently, sometimes by large margins. Code available at https://github.com/NVlabs/GCViT.
PDF Tech. Report
论文截图
Adapting Self-Supervised Vision Transformers by Probing Attention-Conditioned Masking Consistency
Authors:Viraj Prabhu, Sriram Yenamandra, Aaditya Singh, Judy Hoffman
Visual domain adaptation (DA) seeks to transfer trained models to unseen, unlabeled domains across distribution shift, but approaches typically focus on adapting convolutional neural network architectures initialized with supervised ImageNet representations. In this work, we shift focus to adapting modern architectures for object recognition — the increasingly popular Vision Transformer (ViT) — and modern pretraining based on self-supervised learning (SSL). Inspired by the design of recent SSL approaches based on learning from partial image inputs generated via masking or cropping — either by learning to predict the missing pixels, or learning representational invariances to such augmentations — we propose PACMAC, a simple two-stage adaptation algorithm for self-supervised ViTs. PACMAC first performs in-domain SSL on pooled source and target data to learn task-discriminative features, and then probes the model’s predictive consistency across a set of partial target inputs generated via a novel attention-conditioned masking strategy, to identify reliable candidates for self-training. Our simple approach leads to consistent performance gains over competing methods that use ViTs and self-supervised initializations on standard object recognition benchmarks. Code available at https://github.com/virajprabhu/PACMAC
PDF
论文截图
Self-Distilled Self-Supervised Representation Learning
Authors:Jiho Jang, Seonhoon Kim, Kiyoon Yoo, Jangho Kim, Nojun Kwak
State-of-the-art frameworks in self-supervised learning have recently shown that fully utilizing transformer-based models can lead to performance boost compared to conventional CNN models. Striving to maximize the mutual information of two views of an image, existing works apply a contrastive loss to the final representations. Motivated by self-distillation in the supervised regime, we further exploit this by allowing the intermediate representations to learn from the final layer via the contrastive loss. Through self-distillation, the intermediate layers are better suited for instance discrimination, making the performance of an early-exited sub-network not much degraded from that of the full network. This renders the pretext task easier also for the final layer, lead to better representations. Our method, Self-Distilled Self-Supervised Learning (SDSSL), outperforms competitive baselines (SimCLR, BYOL and MoCo v3) using ViT on various tasks and datasets. In the linear evaluation and k-NN protocol, SDSSL not only leads to superior performance in the final layers, but also in most of the lower layers. Furthermore, positive and negative alignments are used to explain how representations are formed more effectively. Code will be available.
PDF 15 pages
论文截图