Recent progress in vision Transformers exhibits great success in various tasks driven by the new spatial modeling mechanism based on dot-product self-attention. In this paper, we show that the key ingredients behind the vision Transformers, namely input-adaptive, long-range and high-order spatial interactions, can also be efficiently implemented with a convolution-based framework. We present the Recursive Gated Convolution ($\textit{g}^\textit{n}$Conv) that performs high-order spatial interactions with gated convolutions and recursive designs. The new operation is highly flexible and customizable, which is compatible with various variants of convolution and extends the two-order interactions in self-attention to arbitrary orders without introducing significant extra computation. $\textit{g}^\textit{n}$Conv can serve as a plug-and-play module to improve various vision Transformers and convolution-based models. Based on the operation, we construct a new family of generic vision backbones named HorNet. Extensive experiments on ImageNet classification, COCO object detection and ADE20K semantic segmentation show HorNet outperform Swin Transformers and ConvNeXt by a significant margin with similar overall architecture and training configurations. HorNet also shows favorable scalability to more training data and larger model sizes. Apart from the effectiveness in visual encoders, we also show $\textit{g}^\textit{n}$Conv can be applied to task-specific decoders and consistently improve dense prediction performance with less computation. Our results demonstrate that $\textit{g}^\textit{n}$Conv can be a new basic module for visual modeling that effectively combines the merits of both vision Transformers and CNNs. Code is available at https://github.com/raoyongming/HorNet
Deep neural networks (DNNs) often have to be compressed, via pruning and/or quantization, before they can be deployed in practical settings. In this work we propose a new compression-aware minimizer dubbed CrAM that modifies the optimization step in a principled way, in order to produce models whose local loss behavior is stable under compression operations such as pruning. Thus, dense models trained via CrAM should be compressible post-training, in a single step, without significant accuracy loss. Experimental results on standard benchmarks, such as residual networks for ImageNet classification and BERT models for language modelling, show that CrAM produces dense models that can be more accurate than the standard SGD/Adam-based baselines, but which are stable under weight pruning: specifically, we can prune models in one-shot to 70-80% sparsity with almost no accuracy loss, and to 90% with reasonable ($\sim 1\%$) accuracy loss, which is competitive with gradual compression methods. Additionally, CrAM can produce sparse models which perform well for transfer learning, and it also works for semi-structured 2:4 pruning patterns supported by GPU hardware. The code for reproducing the results is available at https://github.com/IST-DASLab/CrAM .
The development of accurate methods for multi-label classification (MLC) of remote sensing (RS) images is one of the most important research topics in RS. To address MLC problems, the use of deep neural networks that require a high number of reliable training images annotated by multiple land-cover class labels (multi-labels) has been found popular in RS. However, collecting such annotations is time-consuming and costly. A common procedure to obtain annotations at zero labeling cost is to rely on thematic products or crowdsourced labels. As a drawback, these procedures come with the risk of label noise that can distort the learning process of the MLC algorithms. In the literature, most label noise robust methods are designed for single-label classification (SLC) problems in computer vision (CV), where each image is annotated by a single label. Unlike SLC, label noise in MLC can be associated with: 1) subtractive label-noise (a land cover class label is not assigned to an image while that class is present in the image); 2) additive label-noise (a land cover class label is assigned to an image although that class is not present in the given image); and 3) mixed label-noise (a combination of both). In this paper, we investigate three different noise robust CV SLC methods and adapt them to be robust for multi-label noise scenarios in RS. During experiments, we study the effects of different types of multi-label noise and evaluate the adapted methods rigorously. To this end, we also introduce a synthetic multi-label noise injection strategy that is more adequate to simulate operational scenarios compared to the uniform label noise injection strategy, in which the labels of absent and present classes are flipped at uniform probability. Further, we study the relevance of different evaluation metrics in MLC problems under noisy multi-labels.
Recently, numerous efficient Transformers have been proposed to reduce the quadratic computational complexity of standard Transformers caused by the Softmax attention. However, most of them simply swap Softmax with an efficient attention mechanism without considering the customized architectures specially for the efficient attention. In this paper, we argue that the handcrafted vanilla Transformer architectures for Softmax attention may not be suitable for efficient Transformers. To address this issue, we propose a new framework to find optimal architectures for efficient Transformers with the neural architecture search (NAS) technique. The proposed method is validated on popular machine translation and image classification tasks. We observe that the optimal architecture of the efficient Transformer has the reduced computation compared with that of the standard Transformer, but the general accuracy is less comparable. It indicates that the Softmax attention and efficient attention have their own distinctions but neither of them can simultaneously balance the accuracy and efficiency well. This motivates us to mix the two types of attention to reduce the performance imbalance. Besides the search spaces that commonly used in existing NAS Transformer approaches, we propose a new search space that allows the NAS algorithm to automatically search the attention variants along with architectures. Extensive experiments on WMT' 14 En-De and CIFAR-10 demonstrate that our searched architecture maintains comparable accuracy to the standard Transformer with notably improved computational efficiency.
Memristors, emerging non-volatile memory devices, have shown promising potential in neuromorphic hardware designs, especially in spiking neural network (SNN) hardware implementation. Memristor-based SNNs have been successfully applied in a wide range of various applications, including image classification and pattern recognition. However, implementing memristor-based SNNs in text classification is still under exploration. One of the main reasons is that training memristor-based SNNs for text classification is costly due to the lack of efficient learning rules and memristor non-idealities. To address these issues and accelerate the research of exploring memristor-based spiking neural networks in text classification applications, we develop a simulation framework with a virtual memristor array using an empirical memristor model. We use this framework to demonstrate a sentiment analysis task in the IMDB movie reviews dataset. We take two approaches to obtain trained spiking neural networks with memristor models: 1) by converting a pre-trained artificial neural network (ANN) to a memristor-based SNN, or 2) by training a memristor-based SNN directly. These two approaches can be applied in two scenarios: offline classification and online training. We achieve the classification accuracy of 85.88% by converting a pre-trained ANN to a memristor-based SNN and 84.86% by training the memristor-based SNN directly, given that the baseline training accuracy of the equivalent ANN is 86.02%. We conclude that it is possible to achieve similar classification accuracy in simulation from ANNs to SNNs and from non-memristive synapses to data-driven memristive synapses. We also investigate how global parameters such as spike train length, the read noise, and the weight updating stop conditions affect the neural networks in both approaches.
Convolutional Neural Networks have become the norm in image classification. Nevertheless, their difficulty to maintain high accuracy across datasets has become apparent in the past few years. In order to utilize such models in real-world scenarios and applications, they must be able to provide trustworthy predictions on unseen data. In this paper, we argue that extracting features from a CNN's intermediate layers can assist in the model's final prediction. Specifically, we adapt the Hypercolumns method to a ResNet-18 and find a significant increase in the model's accuracy, when evaluating on the NICO dataset.
Masked image modeling (MIM) has achieved promising results on various vision tasks. However, the limited discriminability of learned representation manifests there is still plenty to go for making a stronger vision learner. Towards this goal, we propose Contrastive Masked Autoencoders (CMAE), a new self-supervised pre-training method for learning more comprehensive and capable vision representations. By elaboratively unifying contrastive learning (CL) and masked image model (MIM) through novel designs, CMAE leverages their respective advantages and learns representations with both strong instance discriminability and local perceptibility. Specifically, CMAE consists of two branches where the online branch is an asymmetric encoder-decoder and the momentum branch is a momentum updated encoder. During training, the online encoder reconstructs original images from latent representations of masked images to learn holistic features. The momentum encoder, fed with the full images, enhances the feature discriminability via contrastive learning with its online counterpart. To make CL compatible with MIM, CMAE introduces two new components, i.e. pixel shifting for generating plausible positive views and feature decoder for complementing features of contrastive pairs. Thanks to these novel designs, CMAE effectively improves the representation quality and transfer performance over its MIM counterpart. CMAE achieves the state-of-the-art performance on highly competitive benchmarks of image classification, semantic segmentation and object detection. Notably, CMAE-Base achieves $85.3\%$ top-1 accuracy on ImageNet and $52.5\%$ mIoU on ADE20k, surpassing previous best results by $0.7\%$ and $1.8\%$ respectively. The source code is publicly accessible at \url{https://github.com/ZhichengHuang/CMAE}.
Clinical diagnosis of the pediatric musculoskeletal system relies on the analysis of medical imaging examinations. In the medical image processing pipeline, semantic segmentation using deep learning algorithms enables an automatic generation of patient-specific three-dimensional anatomical models which are crucial for morphological evaluation. However, the scarcity of pediatric imaging resources may result in reduced accuracy and generalization performance of individual deep segmentation models. In this study, we propose to design a novel multi-task, multi-domain learning framework in which a single segmentation network is optimized over the union of multiple datasets arising from distinct parts of the anatomy. Unlike previous approaches, we simultaneously consider multiple intensity domains and segmentation tasks to overcome the inherent scarcity of pediatric data while leveraging shared features between imaging datasets. To further improve generalization capabilities, we employ a transfer learning scheme from natural image classification, along with a multi-scale contrastive regularization aimed at promoting domain-specific clusters in the shared representations, and multi-joint anatomical priors to enforce anatomically consistent predictions. We evaluate our contributions for performing bone segmentation using three scarce and pediatric imaging datasets of the ankle, knee, and shoulder joints. Our results demonstrate that the proposed approach outperforms individual, transfer, and shared segmentation schemes in Dice metric with statistically sufficient margins. The proposed model brings new perspectives towards intelligent use of imaging resources and better management of pediatric musculoskeletal disorders.
A scanning pixel camera is a novel low-cost, low-power sensor that is not diffraction limited. It produces data as a sequence of samples extracted from various parts of the scene during the course of a scan. It can provide very detailed images at the expense of samplerates and slow image acquisition time. This paper proposes a new algorithm which allows the sensor to adapt the samplerate over the course of this sequence. This makes it possible to overcome some of these limitations by minimising the bandwidth and time required to image and transmit a scene, while maintaining image quality. We examine applications to image classification and semantic segmentation and are able to achieve similar results compared to a fully sampled input, while using 80% fewer samples
Active learning is an important technology for automated machine learning systems. In contrast to Neural Architecture Search (NAS) which aims at automating neural network architecture design, active learning aims at automating training data selection. It is especially critical for training a long-tailed task, in which positive samples are sparsely distributed. Active learning alleviates the expensive data annotation issue through incrementally training models powered with efficient data selection. Instead of annotating all unlabeled samples, it iteratively selects and annotates the most valuable samples. Active learning has been popular in image classification, but has not been fully explored in object detection. Most of current approaches on object detection are evaluated with different settings, making it difficult to fairly compare their performance. To facilitate the research in this field, this paper contributes an active learning benchmark framework named as ALBench for evaluating active learning in object detection. Developed on an automatic deep model training system, this ALBench framework is easy-to-use, compatible with different active learning algorithms, and ensures the same training and testing protocols. We hope this automated benchmark system help researchers to easily reproduce literature's performance and have objective comparisons with prior arts. The code will be release through Github.
Recently, embedding and metric-based few-shot learning (FSL) has been introduced into hyperspectral image classification (HSIC) and achieved impressive progress. To further enhance the performance with few labeled samples, we in this paper propose a novel FSL framework for HSIC with a class-covariance metric (CMFSL). Overall, the CMFSL learns global class representations for each training episode by interactively using training samples from the base and novel classes, and a synthesis strategy is employed on the novel classes to avoid overfitting. During the meta-training and meta-testing, the class labels are determined directly using the Mahalanobis distance measurement rather than an extra classifier. Benefiting from the task-adapted class-covariance estimations, the CMFSL can construct more flexible decision boundaries than the commonly used Euclidean metric. Additionally, a lightweight cross-scale convolutional network (LXConvNet) consisting of 3-D and 2-D convolutions is designed to thoroughly exploit the spectral-spatial information in the high-frequency and low-frequency scales with low computational complexity. Furthermore, we devise a spectral-prior-based refinement module (SPRM) in the initial stage of feature extraction, which can not only force the network to emphasize the most informative bands while suppressing the useless ones, but also alleviate the effects of the domain shift between the base and novel categories to learn a collaborative embedding mapping. Extensive experiment results on four benchmark data sets demonstrate that the proposed CMFSL can outperform the state-of-the-art methods with few-shot annotated samples.
This paper proposes a method for visually explaining the decision-making process of video recognition networks with a temporal extension of occlusion sensitivity analysis, called Adaptive Occlusion Sensitivity Analysis (AOSA). The key idea here is to occlude a specific volume of data by a 3D mask in an input 3D temporal-spatial data space and then measure the change degree in the output score. The occluded volume data that produces a larger change degree is regarded as a more critical element for classification. However, while the occlusion sensitivity analysis is commonly used to analyze single image classification, applying this idea to video classification is not so straightforward as a simple fixed cuboid cannot deal with complicated motions. To solve this issue, we adaptively set the shape of a 3D occlusion mask while referring to motions. Our flexible mask adaptation is performed by considering the temporal continuity and spatial co-occurrence of the optical flows extracted from the input video data. We further propose a novel method to reduce the computational cost of the proposed method with the first-order approximation of the output score with respect to an input video. We demonstrate the effectiveness of our method through various and extensive comparisons with the conventional methods in terms of the deletion/insertion metric and the pointing metric on the UCF101 dataset and the Kinetics-400 and 700 datasets.
Fine-tuning is widely applied in image classification tasks as a transfer learning approach. It re-uses the knowledge from a source task to learn and obtain a high performance in target tasks. Fine-tuning is able to alleviate the challenge of insufficient training data and expensive labelling of new data. However, standard fine-tuning has limited performance in complex data distributions. To address this issue, we propose the Adaptable Multi-tuning method, which adaptively determines each data sample's fine-tuning strategy. In this framework, multiple fine-tuning settings and one policy network are defined. The policy network in Adaptable Multi-tuning can dynamically adjust to an optimal weighting to feed different samples into models that are trained using different fine-tuning strategies. Our method outperforms the standard fine-tuning approach by 1.69%, 2.79% on the datasets FGVC-Aircraft, and Describable Texture, yielding comparable performance on the datasets Stanford Cars, CIFAR-10, and Fashion-MNIST.
Detection of violence and weaponized violence in closed-circuit television (CCTV) footage requires a comprehensive approach. In this work, we introduce the \emph{Smart-City CCTV Violence Detection (SCVD)} dataset, specifically designed to facilitate the learning of weapon distribution in surveillance videos. To tackle the complexities of analyzing 3D surveillance video for violence recognition tasks, we propose a novel technique called \emph{SSIVD-Net} (\textbf{S}alient-\textbf{S}uper-\textbf{I}mage for \textbf{V}iolence \textbf{D}etection). Our method reduces 3D video data complexity, dimensionality, and information loss while improving inference, performance, and explainability through salient-super-Image representations. Considering the scalability and sustainability requirements of futuristic smart cities, the authors introduce the \emph{Salient-Classifier}, a novel architecture combining a kernelized approach with a residual learning strategy. We evaluate variations of SSIVD-Net and Salient Classifier on our SCVD dataset and benchmark against state-of-the-art (SOTA) models commonly employed in violence detection. Our approach exhibits significant improvements in detecting both weaponized and non-weaponized violence instances. By advancing the SOTA in violence detection, our work offers a practical and scalable solution suitable for real-world applications. The proposed methodology not only addresses the challenges of violence detection in CCTV footage but also contributes to the understanding of weapon distribution in smart surveillance. Ultimately, our research findings should enable smarter and more secure cities, as well as enhance public safety measures.
To address the trade-off problem of quality-diversity for the generated images in imbalanced classification tasks, we research on over-sampling based methods at the feature level instead of the data level and focus on searching the latent feature space for optimal distributions. On this basis, we propose an iMproved Estimation Distribution Algorithm based Latent featUre Distribution Evolution (MEDA_LUDE) algorithm, where a joint learning procedure is programmed to make the latent features both optimized and evolved by the deep neural networks and the evolutionary algorithm, respectively. We explore the effect of the Large-margin Gaussian Mixture (L-GM) loss function on distribution learning and design a specialized fitness function based on the similarities among samples to increase diversity. Extensive experiments on benchmark based imbalanced datasets validate the effectiveness of our proposed algorithm, which can generate images with both quality and diversity. Furthermore, the MEDA_LUDE algorithm is also applied to the industrial field and successfully alleviates the imbalanced issue in fabric defect classification.
Single-image haze removal is a long-standing hurdle for computer vision applications. Several works have been focused on transferring advances from image classification, detection, and segmentation to the niche of image dehazing, primarily focusing on contrastive learning and knowledge distillation. However, these approaches prove computationally expensive, raising concern regarding their applicability to on-the-edge use-cases. This work introduces a simple, lightweight, and efficient framework for single-image haze removal, exploiting rich "dark-knowledge" information from a lightweight pre-trained super-resolution model via the notion of heterogeneous knowledge distillation. We designed a feature affinity module to maximize the flow of rich feature semantics from the super-resolution teacher to the student dehazing network. In order to evaluate the efficacy of our proposed framework, its performance as a plug-and-play setup to a baseline model is examined. Our experiments are carried out on the RESIDE-Standard dataset to demonstrate the robustness of our framework to the synthetic and real-world domains. The extensive qualitative and quantitative results provided establish the effectiveness of the framework, achieving gains of upto 15\% (PSNR) while reducing the model size by $\sim$20 times.
Compressive learning (CL) is an emerging framework that integrates signal acquisition via compressed sensing (CS) and machine learning for inference tasks directly on a small number of measurements. It can be a promising alternative to classical image-domain methods and enjoys great advantages in memory saving and computational efficiency. However, previous attempts on CL are not only limited to a fixed CS ratio, which lacks flexibility, but also limited to MNIST/CIFAR-like datasets and do not scale to complex real-world high-resolution (HR) data or vision tasks. In this paper, a novel transformer-based compressive learning framework on large-scale images with arbitrary CS ratios, dubbed TransCL, is proposed. Specifically, TransCL first utilizes the strategy of learnable block-based compressed sensing and proposes a flexible linear projection strategy to enable CL to be performed on large-scale images in an efficient block-by-block manner with arbitrary CS ratios. Then, regarding CS measurements from all blocks as a sequence, a pure transformer-based backbone is deployed to perform vision tasks with various task-oriented heads. Our sufficient analysis presents that TransCL exhibits strong resistance to interference and robust adaptability to arbitrary CS ratios. Extensive experiments for complex HR data demonstrate that the proposed TransCL can achieve state-of-the-art performance in image classification and semantic segmentation tasks. In particular, TransCL with a CS ratio of $10\%$ can obtain almost the same performance as when operating directly on the original data and can still obtain satisfying performance even with an extremely low CS ratio of $1\%$. The source codes of our proposed TransCL is available at \url{https://github.com/MC-E/TransCL/}.
The success of Vision Transformer (ViT) in various computer vision tasks has promoted the ever-increasing prevalence of this convolution-free network. The fact that ViT works on image patches makes it potentially relevant to the problem of jigsaw puzzle solving, which is a classical self-supervised task aiming at reordering shuffled sequential image patches back to their natural form. Despite its simplicity, solving jigsaw puzzle has been demonstrated to be helpful for diverse tasks using Convolutional Neural Networks (CNNs), such as self-supervised feature representation learning, domain generalization, and fine-grained classification. In this paper, we explore solving jigsaw puzzle as a self-supervised auxiliary loss in ViT for image classification, named Jigsaw-ViT. We show two modifications that can make Jigsaw-ViT superior to standard ViT: discarding positional embeddings and masking patches randomly. Yet simple, we find that Jigsaw-ViT is able to improve both in generalization and robustness over the standard ViT, which is usually rather a trade-off. Experimentally, we show that adding the jigsaw puzzle branch provides better generalization than ViT on large-scale image classification on ImageNet. Moreover, the auxiliary task also improves robustness to noisy labels on Animal-10N, Food-101N, and Clothing1M as well as adversarial examples. Our implementation is available at https://yingyichen-cyy.github.io/Jigsaw-ViT/.
In this paper, we propose an encryption method for ConvMixer models with a secret key. Encryption methods for DNN models have been studied to achieve adversarial defense, model protection and privacy-preserving image classification. However, the use of conventional encryption methods degrades the performance of models compared with that of plain models. Accordingly, we propose a novel method for encrypting ConvMixer models. The method is carried out on the basis of an embedding architecture that ConvMixer has, and models encrypted with the method can have the same performance as models trained with plain images only when using test images encrypted with a secret key. In addition, the proposed method does not require any specially prepared data for model training or network modification. In an experiment, the effectiveness of the proposed method is evaluated in terms of classification accuracy and model protection in an image classification task on the CIFAR10 dataset.
Scene inference under low-light is a challenging problem due to severe noise in the captured images. One way to reduce noise is to use longer exposure during the capture. However, in the presence of motion (scene or camera motion), longer exposures lead to motion blur, resulting in loss of image information. This creates a trade-off between these two kinds of image degradations: motion blur (due to long exposure) vs. noise (due to short exposure), also referred as a dual image corruption pair in this paper. With the rise of cameras capable of capturing multiple exposures of the same scene simultaneously, it is possible to overcome this trade-off. Our key observation is that although the amount and nature of degradation varies for these different image captures, the semantic content remains the same across all images. To this end, we propose a method to leverage these multi exposure captures for robust inference under low-light and motion. Our method builds on a feature consistency loss to encourage similar results from these individual captures, and uses the ensemble of their final predictions for robust visual recognition. We demonstrate the effectiveness of our approach on simulated images as well as real captures with multiple exposures, and across the tasks of object detection and image classification.