Vision transformers (ViTs) have found only limited practical use in processing images, in spite of their state-of-the-art accuracy on certain benchmarks. The reason for their limited use include their need for larger training datasets and more computational resources compared to convolutional neural networks (CNNs), owing to the quadratic complexity of their self-attention mechanism. We propose a linear attention-convolution hybrid architecture -- Convolutional X-formers for Vision (CXV) -- to overcome these limitations. We replace the quadratic attention with linear attention mechanisms, such as Performer, Nyströmformer, and Linear Transformer, to reduce its GPU usage. Inductive prior for image data is provided by convolutional sub-layers, thereby eliminating the need for class token and positional embeddings used by the ViTs. We also propose a new training method where we use two different optimizers during different phases of training and show that it improves the top-1 image classification accuracy across different architectures. CXV outperforms other architectures, token mixers (e.g. ConvMixer, FNet and MLP Mixer), transformer models (e.g. ViT, CCT, CvT and hybrid Xformers), and ResNets for image classification in scenarios with limited data and GPU resources (cores, RAM, power).
Although transformers have become the neural architectures of choice for natural language processing, they require orders of magnitude more training data, GPU memory, and computations in order to compete with convolutional neural networks for computer vision. The attention mechanism of transformers scales quadratically with the length of the input sequence, and unrolled images have long sequence lengths. Plus, transformers lack an inductive bias that is appropriate for images. We tested three modifications to vision transformer (ViT) architectures that address these shortcomings. Firstly, we a…
Xformer: Hybrid X-Shaped Transformer for Image Denoising
Shanghai Jiao Tong University（上海交通大学）；Shanghai AI Laboratory（上海人工智能实验室）；University of Sydney（悉尼大学）；Shenyang Institute of Automation, Chinese Academy of Sciences（中国科学院沈阳自动化研究所）；ETH Zürich（苏黎世联邦理工学院）
Artificial and Natural Intelligence Toulouse Institute（图卢兹人工智能与自然智能研究所）；Université de Toulouse（图卢兹大学）；Carney Institute for Brain Science（卡尼脑科学研究所）；Brown University（布朗大学）；Centre de Recherche Cerveau et Cognition（脑与认知研究中心）；CNRS（法国国家科学研究中心）
Vision transformers are nowadays the de-facto choice for image classification tasks. There are two broad categories of classification tasks, fine-grained and coarse-grained. In fine-grained classification, the necessity is to discover subtle differences due to the high level of similarity between sub-classes. Such distinctions are often lost as we downscale the image to save the memory and computational cost associated with vision transformers (ViT). In this work, we present an in-depth analysis and describe the critical components for developing a system for the fine-grained categorization of…
National University of Defense Technology（国防科技大学）；Chinese Academy of Sciences（中国科学院）
Vision transformers have shown excellent performance in computer vision tasks. As the computation cost of their self-attention mechanism is expensive, recent works tried to replace the self-attention mechanism in vision transformers with convolutional operations, which is more efficient with built-in inductive bias. However, these efforts either ignore multi-level features or lack dynamic prosperity, leading to sub-optimal performance. In this paper, we propose a Dynamic Multi-level Attention mechanism (DMA), which captures different patterns of input images by multiple kernel sizes and enable…
Several recent studies have demonstrated that attention-based networks, such as Vision Transformer (ViT), can outperform Convolutional Neural Networks (CNNs) on several computer vision tasks without using convolutional layers. This naturally leads to the following questions: Can a self-attention layer of ViT express any convolution operation? In this work, we prove that a single ViT layer with image patches as the input can perform any convolution operation constructively, where the multi-head attention mechanism and the relative positional encoding play essential roles. We further provide a l…
Ecole Normale Supérieure（巴黎高等师范学院）；Facebook AI Research（脸书人工智能研究院）
Convolutional architectures have proven extremely successful for vision tasks. Their hard inductive biases enable sample-efficient learning, but come at the cost of a potentially lower performance ceiling. Vision Transformers (ViTs) rely on more flexible self-attention layers, and have recently outperformed CNNs for image classification. However, they require costly pre-training on large external datasets or distillation from pre-trained convolutional networks. In this paper, we ask the following question: is it possible to combine the strengths of these two architectures while avoiding their …
Vision transformers rely on a patch token based self attention mechanism, in contrast to convolutional networks. We investigate fundamental differences between these two families of models, by designing a block sparsity based adversarial token attack. We probe and analyze transformer as well as convolutional models with token attacks of varying patch sizes. We infer that transformer models are more sensitive to token attacks than convolutional models, with ResNets outperforming Transformer models by up to $\sim30\%$ in robust accuracy for single token attacks.
University of Science and Technology of China（中国科学技术大学）
There still remains an extreme performance gap between Vision Transformers (ViTs) and Convolutional Neural Networks (CNNs) when training from scratch on small datasets, which is concluded to the lack of inductive bias. In this paper, we further consider this problem and point out two weaknesses of ViTs in inductive biases, that is, the spatial relevance and diverse channel representation. First, on spatial aspect, objects are locally compact and relevant, thus fine-grained feature needs to be extracted from a token and its neighbors. While the lack of data hinders ViTs to attend the spatial re…
École Polytechnique Fédérale de Lausanne (EPFL)（洛桑联邦理工学院）
Recent trends of incorporating attention mechanisms in vision have led researchers to reconsider the supremacy of convolutional layers as a primary building block. Beyond helping CNNs to handle long-range dependencies, Ramachandran et al. (2019) showed that attention can completely replace convolution and achieve state-of-the-art performance on vision tasks. This raises the question: do learned attention layers operate similarly to convolutional layers? This work provides evidence that attention layers can perform convolution and, indeed, they often learn to do so in practice. Specifically, we…
Due to the success of Bidirectional Encoder Representations from Transformers (BERT) in natural language process (NLP), the multi-head attention transformer has been more and more prevalent in computer-vision researches (CV). However, it still remains a challenge for researchers to put forward complex tasks such as vision detection and semantic segmentation. Although multiple Transformer-Based architectures like DETR and ViT-FRCNN have been proposed to complete object detection task, they inevitably decreases discrimination accuracy and brings down computational efficiency caused by the enormo…
In this work we introduce a convolutional neural network (CNN) that jointly handles low-, mid-, and high-level vision tasks in a unified architecture that is trained end-to-end. Such a universal network can act like a `swiss knife' for vision tasks; we call this architecture an UberNet to indicate its overarching nature. We address two main technical challenges that emerge when broadening up the range of tasks handled by a single CNN: (i) training a deep architecture while relying on diverse training sets and (ii) training many (potentially unlimited) tasks with a limited memory budget. Proper…
Deep learning architectures such as convolutional neural networks are the standard in computer vision for image processing tasks. Their accuracy however often comes at the cost of long and computationally expensive training, the need for large annotated datasets, and extensive hyper-parameter searches. On the other hand, a different method known as convolutional extreme learning machine has shown the potential to perform equally with a dramatic decrease in training time. Space imagery, especially about small bodies, could be well suited for this method. In this work, convolutional extreme lear…
College of Computing and Informatics, Drexel University（德雷塞尔大学计算与信息学院）
This paper presents a new concept formation approach that supports the ability to incrementally learn and predict labels for visual images. This work integrates the idea of convolutional image processing, from computer vision research, with a concept formation approach that is based on psychological studies of how humans incrementally form and use concepts. We experimentally evaluate this new approach by applying it to an incremental variation of the MNIST digit recognition task. We compare its performance to Cobweb, a concept formation approach that does not support convolutional processing, …
Convolutional neural networks (CNNs) have so far been the de-facto model for visual data. Recent work has shown that (Vision) Transformer models (ViT) can achieve comparable or even superior performance on image classification tasks. This raises a central question: how are Vision Transformers solving these tasks? Are they acting like convolutional networks, or learning entirely different visual representations? Analyzing the internal representation structure of ViTs and CNNs on image classification benchmarks, we find striking differences between the two architectures, such as ViT having more …
Series expansions have been a cornerstone of applied mathematics and engineering for centuries. In this paper, we revisit the Taylor series expansion from a modern Machine Learning perspective. Specifically, we introduce the Fast Continuous Convolutional Taylor Transform (FC2T2), a variant of the Fast Multipole Method (FMM), that allows for the efficient approximation of low dimensional convolutional operators in continuous space. We build upon the FMM which is an approximate algorithm that reduces the computational complexity of N-body problems from O(NM) to O(N+M) and finds application in e.…
The robust and efficient recognition of visual relations in images is a hallmark of biological vision. We argue that, despite recent progress in visual recognition, modern machine vision algorithms are severely limited in their ability to learn visual relations. Through controlled experiments, we demonstrate that visual-relation problems strain convolutional neural networks (CNNs). The networks eventually break altogether when rote memorization becomes impossible, as when intra-class variability exceeds network capacity. Motivated by the comparable success of biological vision, we argue that f…
OPPO；North China University of Technology（北方工业大学）；Neolix Autonomous Vehicle（新石器自动驾驶）；University of Science and Technology of China（中国科学技术大学）
Vision Transformers (ViTs) have recently dominated a range of computer vision tasks, yet it suffers from low training data efficiency and inferior local semantic representation capability without appropriate inductive bias. Convolutional neural networks (CNNs) inherently capture regional-aware semantics, inspiring researchers to introduce CNNs back into the architecture of the ViTs to provide desirable inductive bias for ViTs. However, is the locality achieved by the micro-level CNNs embedded in ViTs good enough? In this paper, we investigate the problem by profoundly exploring how the macro a…
Traditional architectures for solving computer vision problems and the degree of success they enjoyed have been heavily reliant on hand-crafted features. However, of late, deep learning techniques have offered a compelling alternative -- that of automatically learning problem-specific features. With this new paradigm, every problem in computer vision is now being re-examined from a deep learning perspective. Therefore, it has become important to understand what kind of deep networks are suitable for a given problem. Although general surveys of this fast-moving paradigm (i.e. deep-networks) exi…
We present an approach for adapting convolutional neural networks for object recognition and classification to scientific literature layout detection (SLLD), a shared subtask of several information extraction problems. Scientific publications contain multiple types of information sought by researchers in various disciplines, organized into an abstract, bibliography, and sections documenting related work, experimental methods, and results; however, there is no effective way to extract this information due to their diverse layout. In this paper, we present a novel approach to developing an end-t…
An ever increasing number of computer vision and image/video processing challenges are being approached using deep convolutional neural networks, obtaining state-of-the-art results in object recognition and detection, semantic segmentation, action recognition, optical flow and superresolution. Hardware acceleration of these algorithms is essential to adopt these improvements in embedded and mobile computer vision systems. We present a new architecture, design and implementation as well as the first reported silicon measurements of such an accelerator, outperforming previous work in terms of po…
The pretrain-then-finetune paradigm has been widely adopted in computer vision. But as the size of Vision Transformer (ViT) grows exponentially, the full finetuning becomes prohibitive in view of the heavier storage overhead. Motivated by parameter-efficient transfer learning (PETL) on language transformers, recent studies attempt to insert lightweight adaptation modules (e.g., adapter layers or prompt tokens) to pretrained ViT and only finetune these modules while the pretrained weights are frozen. However, these modules were originally proposed to finetune language models and did not take in…
The growing demand for the internet of things (IoT) makes it necessary to implement computer vision tasks such as object recognition in low-power devices. Convolutional neural networks (CNNs) are a potential approach for object recognition and detection. However, the convolutional layer in CNN consumes significant energy compared to the fully connected layers. To mitigate this problem, a new approach based on the Hadamard transformation as an alternative to the convolution operation is demonstrated using two fundamental datasets, MNIST and CIFAR10. The mathematical expression of the Hadamard m…
The rapid advancement of deep learning models that can generate and synthesis hyper-realistic videos known as Deepfakes and their ease of access to the general public have raised concern from all concerned bodies to their possible malicious intent use. Deep learning techniques can now generate faces, swap faces between two subjects in a video, alter facial expressions, change gender, and alter facial features, to list a few. These powerful video manipulation methods have potential use in many fields. However, they also pose a looming threat to everyone if used for harmful purposes such as iden…
In many computer vision domains, the input images must conform with the pinhole camera model, where straight lines in the real world are projected as straight lines in the image. Performing computer vision tasks on live sports broadcast footage imposes challenging requirements where the algorithms cannot rely on a specific calibration pattern must be able to cope with unknown and uncalibrated cameras, radial distortion originating from complex television lenses, few visual clues to compensate distortion by, and the necessity for real-time performance. We present a novel method for single-image…
F. Hoffmann-La Roche AG（罗氏制药）；Genentech, Inc.（基因泰克公司）
Recently, vision transformers were shown to be capable of outperforming convolutional neural networks when pretrained on sufficient amounts of data. In comparison to convolutional neural networks, vision transformers have a weaker inductive bias and therefore allow a more flexible feature detection. Due to their promising feature detection, this work explores vision transformers for tumor detection in digital pathology whole slide images in four tissue types, and for tissue type identification. We compared the patch-wise classification performance of the vision transformer DeiT-Tiny to the sta…
The puzzle of computer vision might find new challenging solutions when we realize that most successful methods are working at image level, which is remarkably more difficult than processing directly visual streams. In this paper, we claim that their processing naturally leads to formulate the motion invariance principle, which enables the construction of a new theory of learning with convolutional networks. The theory addresses a number of intriguing questions that arise in natural vision, and offers a well-posed computational scheme for the discovery of convolutional filters over the retina.…
In this paper, a method to detect environmental hazards related to a fall risk using a mobile vision system is proposed. First-person perspective videos are proposed to provide objective evidence on cause and circumstances of perturbed balance during activities of daily living, targeted to seniors. A classification problem was defined with 12 total classes of potential fall risks, including slope changes (e.g., stairs, curbs, ramps) and surfaces (e.g., gravel, grass, concrete). Data was collected using a chest-mounted GoPro camera. We developed a convolutional neural network for automatic feat…
Recently, Transformer networks have achieved impressive results on a variety of vision tasks. However, most of them are computationally expensive and not suitable for real-world mobile applications. In this work, we present Mobile Convolutional Vision Transformer (MoCoViT), which improves in performance and efficiency by introducing transformer into mobile convolutional networks to leverage the benefits of both architectures. Different from recent works on vision transformer, the mobile transformer block in MoCoViT is carefully designed for mobile devices and is very lightweight, accomplished …
University of Oregon（俄勒冈大学）；UIUC（伊利诺伊大学厄巴纳-香槟分校）；Picsart AI Research (PAIR)（Picsart人工智能研究院）
MLP-based architectures, which consist of a sequence of consecutive multi-layer perceptron blocks, have recently been found to reach comparable results to convolutional and transformer-based methods. However, most adopt spatial MLPs which take fixed dimension inputs, therefore making it difficult to apply them to downstream tasks, such as object detection and semantic segmentation. Moreover, single-stage designs further limit performance in other computer vision tasks and fully connected layers bear heavy computation. To tackle these problems, we propose ConvMLP: a hierarchical Convolutional M…
Nanjing University of Science and Technology（南京理工大学）；NVIDIA（英伟达）；City University of Hong Kong（香港城市大学）；Dalian University of Technology（大连理工大学）；SenseTime Research（商汤科技研究院）；Tencent Youtu Lab（腾讯优图实验室）；UC Merced（加州大学默塞德分校）
In this paper, we propose a general dual convolutional neural network (DualCNN) for low-level vision problems, e.g., super-resolution, edge-preserving filtering, deraining and dehazing. These problems usually involve the estimation of two components of the target signals: structures and details. Motivated by this, our proposed DualCNN consists of two parallel branches, which respectively recovers the structures and details in an end-to-end manner. The recovered structures and details can generate the target signals according to the formation model for each particular application. The DualCNN i…
PDPM Indian Institute of Information Technology, Design and Manufacturing（PDPM印度信息技术设计与制造学院）
Plant diseases are the primary cause of crop losses globally, with an impact on the world economy. To deal with these issues, smart agriculture solutions are evolving that combine the Internet of Things and machine learning for early disease detection and control. Many such systems use vision-based machine learning methods for real-time disease detection and diagnosis. With the advancement in deep learning techniques, new methods have emerged that employ convolutional neural networks for plant disease detection and identification. Another trend in vision-based deep learning is the use of visio…
University of Illinois at Urbana-Champaign（伊利诺伊大学厄巴纳-香槟分校）
Current methods of practice for inspection of civil infrastructure typically involve visual assessments conducted manually by trained inspectors. For post-earthquake structural inspections, the number of structures to be inspected often far exceeds the capability of the available inspectors. The labor intensive and time consuming natures of manual inspection have engendered research into development of algorithms for automated damage identification using computer vision techniques. In this paper, a novel damage localization and classification technique based on a state of the art computer visi…
Detection of pulmonary pathologies using convolutional neural networks, Data Augmentation, ResNet50 and Vision Transformers
Universidad Abierta Interamericana（泛美开放大学）；Faculty of Information Technology（信息技术学院）；Center for Advanced Studies in Information Technology（信息技术高级研究中心）；Interamerican Open University（泛美开放大学）
University of Sydney（悉尼大学）；Huawei Noah’s Ark Lab（华为诺亚方舟实验室）
Vision transformers have been successfully applied to image recognition tasks due to their ability to capture long-range dependencies within an image. However, there are still gaps in both performance and computational cost between transformers and existing convolutional neural networks (CNNs). In this paper, we aim to address this issue and develop a network that can outperform not only the canonical transformers, but also the high-performance convolutional models. We propose a new transformer based hybrid network by taking advantage of transformers to capture long-range dependencies, and of …
We present in this paper a new architecture, named Convolutional vision Transformer (CvT), that improves Vision Transformer (ViT) in performance and efficiency by introducing convolutions into ViT to yield the best of both designs. This is accomplished through two primary modifications: a hierarchy of Transformers containing a new convolutional token embedding, and a convolutional Transformer block leveraging a convolutional projection. These changes introduce desirable properties of convolutional neural networks (CNNs) to the ViT architecture (\ie shift, scale, and distortion invariance) whil…
IBM Research（IBM研究院）；SynSense（SynSense（芯仑科技））；ETH Zürich（苏黎世联邦理工学院）；University of Zürich（苏黎世大学）
Event-based dynamic vision sensors provide very sparse output in the form of spikes, which makes them suitable for low-power applications. Convolutional spiking neural networks model such event-based data and develop their full energy-saving potential when deployed on asynchronous neuromorphic hardware. Event-based vision being a nascent field, the sensitivity of spiking neural networks to potentially malicious adversarial attacks has received little attention so far. We show how white-box adversarial attack algorithms can be adapted to the discrete and sparse nature of event-based visual data…
Universitat Pompeu Fabra（庞培法布拉大学）；University of Adelaide（阿德莱德大学）；ICREA（加泰罗尼亚研究与创新高级研究所）
This paper compares well-established Convolutional Neural Networks (CNNs) to recently introduced Vision Transformers for the task of Diabetic Foot Ulcer Classification, in the context of the DFUC 2021 Grand-Challenge, in which this work attained the first position. Comprehensive experiments demonstrate that modern CNNs are still capable of outperforming Transformers in a low-data regime, likely owing to their ability for better exploiting spatial correlations. In addition, we empirically demonstrate that the recent Sharpness-Aware Minimization (SAM) optimization algorithm considerably improves…
We present a class of efficient models called MobileNets for mobile and embedded vision applications. MobileNets are based on a streamlined architecture that uses depth-wise separable convolutions to build light weight deep neural networks. We introduce two simple global hyper-parameters that efficiently trade off between latency and accuracy. These hyper-parameters allow the model builder to choose the right sized model for their application based on the constraints of the problem. We present extensive experiments on resource and accuracy tradeoffs and show strong performance compared to othe…
