Vision Transformer

2022-09-21 更新

Can We Solve 3D Vision Tasks Starting from A 2D Vision Transformer?

Authors:Yi Wang, Zhiwen Fan, Tianlong Chen, Hehe Fan, Zhangyang Wang

Vision Transformers (ViTs) have proven to be effective, in solving 2D image understanding tasks by training over large-scale image datasets; and meanwhile as a somehow separate track, in modeling the 3D visual world too such as voxels or point clouds. However, with the growing hope that transformers can become the “universal” modeling tool for heterogeneous data, ViTs for 2D and 3D tasks have so far adopted vastly different architecture designs that are hardly transferable. That invites an (over-)ambitious question: can we close the gap between the 2D and 3D ViT architectures? As a piloting study, this paper demonstrates the appealing promise to understand the 3D visual world, using a standard 2D ViT architecture, with only minimal customization at the input and output levels without redesigning the pipeline. To build a 3D ViT from its 2D sibling, we “inflate” the patch embedding and token sequence, accompanied with new positional encoding mechanisms designed to match the 3D data geometry. The resultant “minimalist” 3D ViT, named Simple3D-Former, performs surprisingly robustly on popular 3D tasks such as object classification, point cloud segmentation and indoor scene detection, compared to highly customized 3D-specific designs. It can hence act as a strong baseline for new 3D ViTs. Moreover, we note that pursing a unified 2D-3D ViT design has practical relevance besides just scientific curiosity. Specifically, we demonstrate that Simple3D-Former naturally enables to exploit the wealth of pre-trained weights from large-scale realistic 2D images (e.g., ImageNet), which can be plugged in to enhancing the 3D task performance “for free”.


HiMFR: A Hybrid Masked Face Recognition Through Face Inpainting

Authors:Md Imran Hosen, Md Baharul Islam

To recognize the masked face, one of the possible solutions could be to restore the occluded part of the face first and then apply the face recognition method. Inspired by the recent image inpainting methods, we propose an end-to-end hybrid masked face recognition system, namely HiMFR, consisting of three significant parts: masked face detector, face inpainting, and face recognition. The masked face detector module applies a pretrained Vision Transformer (ViT_b32) to detect whether faces are covered with masked or not. The inpainting module uses a fine-tune image inpainting model based on a Generative Adversarial Network (GAN) to restore faces. Finally, the hybrid face recognition module based on ViT with an EfficientNetB3 backbone recognizes the faces. We have implemented and evaluated our proposed method on four different publicly available datasets: CelebA, SSDMNV2, MAFA, {Pubfig83} with our locally collected small dataset, namely Face5. Comprehensive experimental results show the efficacy of the proposed HiMFR method with competitive performance. Code is available at
PDF 7 pages, 6 figures, International Conference on Pattern Recognition Workshop: Deep Learning for Visual Detection and Recognition


Sequencer: Deep LSTM for Image Classification

Authors:Yuki Tatsunami, Masato Taki

In recent computer vision research, the advent of the Vision Transformer (ViT) has rapidly revolutionized various architectural design efforts: ViT achieved state-of-the-art image classification performance using self-attention found in natural language processing, and MLP-Mixer achieved competitive performance using simple multi-layer perceptrons. In contrast, several studies have also suggested that carefully redesigned convolutional neural networks (CNNs) can achieve advanced performance comparable to ViT without resorting to these new ideas. Against this background, there is growing interest in what inductive bias is suitable for computer vision. Here we propose Sequencer, a novel and competitive architecture alternative to ViT that provides a new perspective on these issues. Unlike ViTs, Sequencer models long-range dependencies using LSTMs rather than self-attention layers. We also propose a two-dimensional version of Sequencer module, where an LSTM is decomposed into vertical and horizontal LSTMs to enhance performance. Despite its simplicity, several experiments demonstrate that Sequencer performs impressively well: Sequencer2D-L, with 54M parameters, realizes 84.6% top-1 accuracy on only ImageNet-1K. Not only that, we show that it has good transferability and the robust resolution adaptability on double resolution-band.
PDF Accepted in NeurIPS 2022; rebuttal edition. Correction of the throughput and adding additional experiments, etc


Attentive Symmetric Autoencoder for Brain MRI Segmentation

Authors:Junjia Huang, Haofeng Li, Guanbin Li, Xiang Wan

Self-supervised learning methods based on image patch reconstruction have witnessed great success in training auto-encoders, whose pre-trained weights can be transferred to fine-tune other downstream tasks of image understanding. However, existing methods seldom study the various importance of reconstructed patches and the symmetry of anatomical structures, when they are applied to 3D medical images. In this paper we propose a novel Attentive Symmetric Auto-encoder (ASA) based on Vision Transformer (ViT) for 3D brain MRI segmentation tasks. We conjecture that forcing the auto-encoder to recover informative image regions can harvest more discriminative representations, than to recover smooth image patches. Then we adopt a gradient based metric to estimate the importance of each image patch. In the pre-training stage, the proposed auto-encoder pays more attention to reconstruct the informative patches according to the gradient metrics. Moreover, we resort to the prior of brain structures and develop a Symmetric Position Encoding (SPE) method to better exploit the correlations between long-range but spatially symmetric regions to obtain effective features. Experimental results show that our proposed attentive symmetric auto-encoder outperforms the state-of-the-art self-supervised learning methods and medical image segmentation models on three brain MRI segmentation benchmarks.
PDF MICCAI 2022, code:


A Vision Transformer-Based Approach to Bearing Fault Classification via Vibration Signals

Authors:Abid Hasan Zim, Aeyan Ashraf, Aquib Iqbal, Asad Malik, Minoru Kuribayashi

Rolling bearings are the most crucial components of rotating machinery. Identifying defective bearings in a timely manner may prevent the malfunction of an entire machinery system. The mechanical condition monitoring field has entered the big data phase as a result of the fast advancement of machine parts. When working with large amounts of data, the manual feature extraction approach has the drawback of being inefficient and inaccurate. Data-driven methods like the Deep Learning method have been successfully used in recent years for mechanical intelligent fault detection. Convolutional neural networks (CNNs) were mostly used in earlier research to detect and identify bearing faults. The CNN model, however, suffers from the drawback of having trouble managing fault-time information, which results in a lack of classification results. In this study, bearing defects have been classified using a state-of-the-art Vision Transformer (ViT). Bearing defects were classified using Case Western Reserve University (CWRU) bearing failure laboratory experimental data. The research took into account 13 distinct kinds of defects under 0-load situations in addition to normal bearing conditions. Using the short-time Fourier transform (STFT), the vibration signals were converted into 2D time-frequency images. The 2D time-frequency images are used as input parameters for the ViT. The model achieved an overall accuracy of 98.8%.


Fast Vision Transformers with HiLo Attention

Authors:Zizheng Pan, Jianfei Cai, Bohan Zhuang

Vision Transformers (ViTs) have triggered the most recent and significant breakthroughs in computer vision. Their efficient designs are mostly guided by the indirect metric of computational complexity, i.e., FLOPs, which however has a clear gap with the direct metric such as throughput. Thus, we propose to use the direct speed evaluation on the target platform as the design principle for efficient ViTs. Particularly, we introduce LITv2, a simple and effective ViT which performs favourably against the existing state-of-the-art methods across a spectrum of different model sizes with faster speed. At the core of LITv2 is a novel self-attention mechanism, which we dub HiLo. HiLo is inspired by the insight that high frequencies in an image capture local fine details and low frequencies focus on global structures, whereas a multi-head self-attention layer neglects the characteristic of different frequencies. Therefore, we propose to disentangle the high/low frequency patterns in an attention layer by separating the heads into two groups, where one group encodes high frequencies via self-attention within each local window, and another group performs the attention to model the global relationship between the average-pooled low-frequency keys from each window and each query position in the input feature map. Benefiting from the efficient design for both groups, we show that HiLo is superior to the existing attention mechanisms by comprehensively benchmarking FLOPs, speed and memory consumption on GPUs. Powered by HiLo, LITv2 serves as a strong backbone for mainstream vision tasks including image classification, dense detection and segmentation. Code is available at
PDF Accepted to NeurIPS 2022


ConvFormer: Closing the Gap Between CNN and Vision Transformers

Authors:Zimian Wei, Hengyue Pan, Xin Niu, Dongsheng Li

Vision transformers have shown excellent performance in computer vision tasks. However, the computation cost of their (local) self-attention mechanism is expensive. Comparatively, CNN is more efficient with built-in inductive bias. Recent works show that CNN is promising to compete with vision transformers by learning their architecture design and training protocols. Nevertheless, existing methods either ignore multi-level features or lack dynamic prosperity, leading to sub-optimal performance. In this paper, we propose a novel attention mechanism named MCA, which captures different patterns of input images by multiple kernel sizes and enables input-adaptive weights with a gating mechanism. Based on MCA, we present a neural network named ConvFormer. ConvFormer adopts the general architecture of vision transformers, while replacing the (local) self-attention mechanism with our proposed MCA. Extensive experimental results demonstrated that ConvFormer outperforms similar size vision transformers(ViTs) and convolutional neural networks (CNNs) in various tasks. For example, ConvFormer-S, ConvFormer-L achieve state-of-the-art performance of 82.8%, 83.6% top-1 accuracy on ImageNet dataset. Moreover, ConvFormer-S outperforms Swin-T by 1.5 mIoU on ADE20K, and 0.9 bounding box AP on COCO with a smaller model size. Code and models will be available.


PaLI: A Jointly-Scaled Multilingual Language-Image Model

Authors:Xi Chen, Xiao Wang, Soravit Changpinyo, AJ Piergiovanni, Piotr Padlewski, Daniel Salz, Sebastian Goodman, Adam Grycner, Basil Mustafa, Lucas Beyer, Alexander Kolesnikov, Joan Puigcerver, Nan Ding, Keran Rong, Hassan Akbari, Gaurav Mishra, Linting Xue, Ashish Thapliyal, James Bradbury, Weicheng Kuo, Mojtaba Seyedhosseini, Chao Jia, Burcu Karagol Ayan, Carlos Riquelme, Andreas Steiner, Anelia Angelova, Xiaohua Zhai, Neil Houlsby, Radu Soricut

Effective scaling and a flexible task interface enable large language models to excel at many tasks. PaLI (Pathways Language and Image model) extends this approach to the joint modeling of language and vision. PaLI generates text based on visual and textual inputs, and with this interface performs many vision, language, and multimodal tasks, in many languages. To train PaLI, we make use of large pretrained encoder-decoder language models and Vision Transformers (ViTs). This allows us to capitalize on their existing capabilities and leverage the substantial cost of training them. We find that joint scaling of the vision and language components is important. Since existing Transformers for language are much larger than their vision counterparts, we train the largest ViT to date (ViT-e) to quantify the benefits from even larger-capacity vision models. To train PaLI, we create a large multilingual mix of pretraining tasks, based on a new image-text training set containing 10B images and texts in over 100 languages. PaLI achieves state-of-the-art in multiple vision and language tasks (such as captioning, visual question-answering, scene-text understanding), while retaining a simple, modular, and scalable design.


文章作者: 木子已
版权声明: 本博客所有文章除特別声明外,均采用 CC BY 4.0 许可协议。转载请注明来源 木子已 !