Unsupervised contrastive learning achieves great success in learning image representations with CNN. Unlike most recent methods that focused on improving accuracy of image classification, we present a novel contrastive learning approach, named DetCo, which fully explores the contrasts between global image and local image patches to learn discriminative representations for object detection. DetCo has several appealing benefits. (1) It is carefully designed by investigating the weaknesses of current self-supervised methods, which discard important representations for object detection. (2) DetCo builds hierarchical intermediate contrastive losses between global image and local patches to improve object detection, while maintaining global representations for image recognition. Theoretical analysis shows that the local patches actually remove the contextual information of an image, improving the lower bound of mutual information for better contrastive learning. (3) Extensive experiments on PASCAL VOC, COCO and Cityscapes demonstrate that DetCo not only outperforms state-of-the-art methods on object detection, but also on segmentation, pose estimation, and 3D shape prediction, while it is still competitive on image classification. For example, on PASCAL VOC, DetCo-100ep achieves 57.4 mAP, which is on par with the result of MoCov2-800ep. Moreover, DetCo consistently outperforms supervised method by 1.6/1.2/1.0 AP on Mask RCNN-C4/FPN/RetinaNet with 1x schedule. Code will be released at \href{https://github.com/xieenze/DetCo}{\color{blue}{\tt github.com/xieenze/DetCo}}.
For visual object recognition tasks, the illumination variations can cause distinct changes in object appearance and thus confuse the deep neural network based recognition models. Especially for some rare illumination conditions, collecting sufficient training samples could be time-consuming and expensive. To solve this problem, in this paper we propose a novel neural network architecture called Separating-Illumination Network (Sill-Net). Sill-Net learns to separate illumination features from images, and then during training we augment training samples with these separated illumination features in the feature space. Experimental results demonstrate that our approach outperforms current state-of-the-art methods in several object classification benchmarks.
A fundamental limitation of applying semi-supervised learning in real-world settings is the assumption that unlabeled test data contains only classes previously encountered in the labeled training data. However, this assumption rarely holds for data in-the-wild, where instances belonging to novel classes may appear at testing time. Here, we introduce a novel open-world semi-supervised learning setting that formalizes the notion that novel classes may appear in the unlabeled test data. In this novel setting, the goal is to solve the class distribution mismatch between labeled and unlabeled data, where at the test time every input instance either needs to be classified into one of the existing classes or a new unseen class needs to be initialized. To tackle this challenging problem, we propose ORCA, an end-to-end deep learning approach that introduces uncertainty adaptive margin mechanism to circumvent the bias towards seen classes caused by learning discriminative features for seen classes faster than for the novel classes. In this way, ORCA reduces the gap between intra-class variance of seen with respect to novel classes. Experiments on image classification datasets and a single-cell annotation dataset demonstrate that ORCA consistently outperforms alternative baselines, achieving 25% improvement on seen and 96% improvement on novel classes of the ImageNet dataset.
State-of-the-art deep-learning systems use decision rules that are challenging for humans to model. Explainable AI (XAI) attempts to improve human understanding but rarely accounts for how people typically reason about unfamiliar agents. We propose explicitly modeling the human explainee via Bayesian Teaching, which evaluates explanations by how much they shift explainees' inferences toward a desired goal. We assess Bayesian Teaching in a binary image classification task across a variety of contexts. Absent intervention, participants predict that the AI's classifications will match their own, but explanations generated by Bayesian Teaching improve their ability to predict the AI's judgements by moving them away from this prior belief. Bayesian Teaching further allows each case to be broken down into sub-examples (here saliency maps). These sub-examples complement whole examples by improving error detection for familiar categories, whereas whole examples help predict correct AI judgements of unfamiliar cases.
State-of-the-art object detectors are vulnerable to localized patch hiding attacks, where an adversary introduces a small adversarial patch to make detectors miss the detection of salient objects. The patch attacker can carry out a physical-world attack by printing and attaching an adversarial patch to the victim object. In this paper, we propose DetectorGuard as the first general framework for building provably robust object detectors against localized patch hiding attacks. DetectorGuard is inspired by recent advancements in robust image classification research; we ask: can we adapt robust image classifiers for robust object detection? Unfortunately, due to their task difference, an object detector naively adapted from a robust image classifier 1) may not necessarily be robust in the adversarial setting or 2) even maintain decent performance in the clean setting. To build a high-performance robust object detector, we propose an objectness explaining strategy: we adapt a robust image classifier to predict objectness for every image location and then explain each objectness using the bounding boxes predicted by a conventional object detector. If all objectness is well explained, we output the predictions made by the conventional object detector; otherwise, we issue an attack alert. Notably, 1) in the adversarial setting, we formally prove the end-to-end robustness of DetectorGuard on certified objects, i.e., it either detects the object or triggers an alert, against any patch hiding attacker within our threat model; 2) in the clean setting, we have almost the same performance as state-of-the-art object detectors. Our evaluation on the PASCAL VOC, MS COCO, and KITTI datasets further demonstrates that DetectorGuard achieves the first provable robustness against localized patch hiding attacks at a negligible cost (<1%) of clean performance.
Convolutional neural networks (CNNs) are able to attain better visual recognition performance than fully connected neural networks despite having much fewer parameters due to their parameter sharing principle. Modern architectures usually contain a small number of fully-connected layers, often at the end, after multiple layers of convolutions. In some cases, most of the convolutions can be eliminated without suffering any loss in recognition performance. However, there is no solid recipe to detect the hidden subset of convolutional neurons that is responsible for the majority of the recognition work. In this work, we formulate this as a pruning problem where the aim is to prune as many kernels as possible while preserving the vanilla generalization performance. To this end, we use the matrix characteristics based on eigenvalues for pruning, in comparison to the average absolute weight of a kernel which is the de facto standard in the literature to assess the importance of an individual convolutional kernel, to shed light on the internal mechanisms of a widely used family of CNNs, namely residual neural networks (ResNets), for the image classification problem using CIFAR-10, CIFAR-100 and Tiny ImageNet datasets.
Convolutional Neural Networks (CNNs) have proven to be a powerful state-of-the-art method for image classification tasks. One drawback however is the high computational complexity and high memory consumption of CNNs which makes them unfeasible for execution on embedded platforms which are constrained on physical resources needed to support CNNs. Quantization has often been used to efficiently optimize CNNs for memory and computational complexity at the cost of a loss of prediction accuracy. We therefore propose a method to optimally quantize the weights, biases and activations of each layer of a pre-trained CNN while controlling the loss in inference accuracy to enable quantized inference. We quantize the 32-bit floating-point precision parameters to low bitwidth fixed-point representations thereby finding optimal bitwidths and fractional offsets for parameters of each layer of a given CNN. We quantize parameters of a CNN post-training without re-training it. Our method is designed to quantize parameters of a CNN taking into account how other parameters are quantized because ignoring quantization errors due to other quantized parameters leads to a low precision CNN with accuracy losses of up to 50% which is far beyond what is acceptable. Our final method therefore gives a low precision CNN with accuracy losses of less than 1%. As compared to a method used by commercial tools that quantize all parameters to 8-bits, our approach provides quantized CNN with averages of 53% lower memory consumption and 77.5% lower cost of executing multiplications for the two CNNs trained on the four datasets that we tested our work on. We find that layer-wise quantization of parameters significantly helps in this process.
The ability of a deep learning network to distinguish between in-distribution (ID) and out-of-distribution (OOD) inputs is crucial for ensuring the reliability and trustworthiness of AI systems. Existing OOD detection methods often involve complex architectural innovations, such as ensemble models, which, while enhancing detection accuracy, significantly increase model complexity and training time. Other methods utilize surrogate samples to simulate OOD inputs, but these may not generalize well across different types of OOD data. In this paper, we propose a straightforward yet novel technique to enhance OOD detection in pre-trained networks without altering its original parameters. Our approach defines probabilistic trust intervals for each network weight, determined using in-distribution data. During inference, additional weight values are sampled, and the resulting disagreements among outputs are utilized for OOD detection. We propose a metric to quantify this disagreement and validate its effectiveness with empirical evidence. Our method significantly outperforms various baseline methods across multiple OOD datasets without requiring actual or surrogate OOD samples. We evaluate our approach on MNIST, Fashion-MNIST, CIFAR-10, CIFAR-100 and CIFAR-10-C (a corruption-augmented version of CIFAR-10), across various neural network architectures (e.g., VGG-16, ResNet-20, DenseNet-100). On the MNIST-FashionMNIST setup, our method achieves a False Positive Rate (FPR) of 12.46\% at 95\% True Positive Rate (TPR), compared to 27.09\% achieved by the best baseline. On adversarial and corrupted datasets such as CIFAR-10-C, our proposed method easily differentiate between clean and noisy inputs. These results demonstrate the robustness of our approach in identifying corrupted and adversarial inputs, all without requiring OOD samples during training.
Fine-grained visual classification is a challenging task that recognizes the sub-classes belonging to the same meta-class. Large inter-class similarity and intra-class variance is the main challenge of this task. Most exiting methods try to solve this problem by designing complex model structures to explore more minute and discriminative regions. In this paper, we argue that mining multi-regional multi-grained features is precisely the key to this task. Specifically, we introduce a new loss function, termed top-down spatial attention loss (TDSA-Loss), which contains a multi-stage channel constrained module and a top-down spatial attention module. The multi-stage channel constrained module aims to make the feature channels in different stages category-aligned. Meanwhile, the top-down spatial attention module uses the attention map generated by high-level aligned feature channels to make middle-level aligned feature channels to focus on particular regions. Finally, we can obtain multiple discriminative regions on high-level feature channels and obtain multiple more minute regions within these discriminative regions on middle-level feature channels. In summary, we obtain multi-regional multi-grained features. Experimental results over four widely used fine-grained image classification datasets demonstrate the effectiveness of the proposed method. Ablative studies further show the superiority of two modules in the proposed method. Codes are available at: https://github.com/dongliangchang/Top-Down-Spatial-Attention-Loss.
Recently, unsupervised local learning, based on Hebb's idea that change in synaptic efficacy depends on the activity of the pre- and postsynaptic neuron only, has shown potential as an alternative training mechanism to backpropagation. Unfortunately, Hebbian learning remains experimental and rarely makes it way into standard deep learning frameworks. In this work, we investigate the potential of Hebbian learning in the context of standard deep learning workflows. To this end, a framework for thorough and systematic evaluation of local learning rules in existing deep learning pipelines is proposed. Using this framework, the potential of Hebbian learned feature extractors for image classification is illustrated. In particular, the framework is used to expand the Krotov-Hopfield learning rule to standard convolutional neural networks without sacrificing accuracy compared to end-to-end backpropagation. The source code is available at https://github.com/Joxis/pytorch-hebbian.
Accuracy predictor is a key component in Neural Architecture Search (NAS) for ranking architectures. Building a high-quality accuracy predictor usually costs enormous computation. To address this issue, instead of using an accuracy predictor, we propose a novel zero-shot index dubbed Zen-Score to rank the architectures. The Zen-Score represents the network expressivity and positively correlates with the model accuracy. The calculation of Zen-Score only takes a few forward inferences through a randomly initialized network, without training network parameters. Built upon the Zen-Score, we further propose a new NAS algorithm, termed as Zen-NAS, by maximizing the Zen-Score of the target network under given inference budgets. Within less than half GPU day, Zen-NAS is able to directly search high performance architectures in a data-free style. Comparing with previous NAS methods, the proposed Zen-NAS is magnitude times faster on multiple server-side and mobile-side GPU platforms with state-of-the-art accuracy on ImageNet. Our source code and pre-trained models are released on https://github.com/idstcv/ZenNAS.
Deep convolutional neural networks, assisted by architectural design strategies, make extensive use of data augmentation techniques and layers with a high number of feature maps to embed object transformations. That is highly inefficient and for large datasets implies a massive redundancy of features detectors. Even though capsules networks are still in their infancy, they constitute a promising solution to extend current convolutional networks and endow artificial visual perception with a process to encode more efficiently all feature affine transformations. Indeed, a properly working capsule network should theoretically achieve higher results with a considerably lower number of parameters count due to intrinsic capability to generalize to novel viewpoints. Nevertheless, little attention has been given to this relevant aspect. In this paper, we investigate the efficiency of capsule networks and, pushing their capacity to the limits with an extreme architecture with barely 160K parameters, we prove that the proposed architecture is still able to achieve state-of-the-art results on three different datasets with only 2% of the original CapsNet parameters. Moreover, we replace dynamic routing with a novel non-iterative, highly parallelizable routing algorithm that can easily cope with a reduced number of capsules. Extensive experimentation with other capsule implementations has proved the effectiveness of our methodology and the capability of capsule networks to efficiently embed visual representations more prone to generalization.
Transformers, which are popular for language modeling, have been explored for solving vision tasks recently, e.g., the Vision Transformer (ViT) for image classification. The ViT model splits each image into a sequence of tokens with fixed length and then applies multiple Transformer layers to model their global relation for classification. However, ViT achieves inferior performance to CNNs when trained from scratch on a midsize dataset like ImageNet. We find it is because: 1) the simple tokenization of input images fails to model the important local structure such as edges and lines among neighboring pixels, leading to low training sample efficiency; 2) the redundant attention backbone design of ViT leads to limited feature richness for fixed computation budgets and limited training samples. To overcome such limitations, we propose a new Tokens-To-Token Vision Transformer (T2T-ViT), which incorporates 1) a layer-wise Tokens-to-Token (T2T) transformation to progressively structurize the image to tokens by recursively aggregating neighboring Tokens into one Token (Tokens-to-Token), such that local structure represented by surrounding tokens can be modeled and tokens length can be reduced; 2) an efficient backbone with a deep-narrow structure for vision transformer motivated by CNN architecture design after empirical study. Notably, T2T-ViT reduces the parameter count and MACs of vanilla ViT by half, while achieving more than 3.0\% improvement when trained from scratch on ImageNet. It also outperforms ResNets and achieves comparable performance with MobileNets by directly training on ImageNet. For example, T2T-ViT with comparable size to ResNet50 (21.5M parameters) can achieve 83.3\% top1 accuracy in image resolution 384$\times$384 on ImageNet. (Code: https://github.com/yitu-opensource/T2T-ViT)
Multi-label zero-shot learning strives to classify images into multiple unseen categories for which no data is available during training. The test samples can additionally contain seen categories in the generalized variant. Existing approaches rely on learning either shared or label-specific attention from the seen classes. Nevertheless, computing reliable attention maps for unseen classes during inference in a multi-label setting is still a challenge. In contrast, state-of-the-art single-label generative adversarial network (GAN) based approaches learn to directly synthesize the class-specific visual features from the corresponding class attribute embeddings. However, synthesizing multi-label features from GANs is still unexplored in the context of zero-shot setting. In this work, we introduce different fusion approaches at the attribute-level, feature-level and cross-level (across attribute and feature-levels) for synthesizing multi-label features from their corresponding multi-label class embedding. To the best of our knowledge, our work is the first to tackle the problem of multi-label feature synthesis in the (generalized) zero-shot setting. Comprehensive experiments are performed on three zero-shot image classification benchmarks: NUS-WIDE, Open Images and MS COCO. Our cross-level fusion-based generative approach outperforms the state-of-the-art on all three datasets. Furthermore, we show the generalization capabilities of our fusion approach in the zero-shot detection task on MS COCO, achieving favorable performance against existing methods. The source code is available at https://github.com/akshitac8/Generative_MLZSL.
Recently demonstrated physical-world adversarial attacks have exposed vulnerabilities in perception systems that pose severe risks for safety-critical applications such as autonomous driving. These attacks place adversarial artifacts in the physical world that indirectly cause the addition of a universal patch to inputs of a model that can fool it in a variety of contexts. Adversarial training is the most effective defense against image-dependent adversarial attacks. However, tailoring adversarial training to universal patches is computationally expensive since the optimal universal patch depends on the model weights which change during training. We propose meta adversarial training (MAT), a novel combination of adversarial training with meta-learning, which overcomes this challenge by meta-learning universal patches along with model training. MAT requires little extra computation while continuously adapting a large set of patches to the current model. MAT considerably increases robustness against universal patch attacks on image classification and traffic-light detection.
This concept paper aims to provide a brief outline of quantum computers, explore existing methods of quantum image classification techniques, so focusing on remote sensing applications, and discuss the bottlenecks of performing these algorithms on currently available open source platforms. Initial results demonstrate feasibility. Next steps include expanding the size of the quantum hidden layer and increasing the variety of output image options.
We propose a novel method to detect and visualize malware through image classification. The executable binaries are represented as grayscale images obtained from the count of N-grams (N=2) of bytes in the Discrete Cosine Transform (DCT) domain and a neural network is trained for malware detection. A shallow neural network is trained for classification, and its accuracy is compared with deep-network architectures such as ResNet that are trained using transfer learning. Neither dis-assembly nor behavioral analysis of malware is required for these methods. Motivated by the visual similarity of these images for different malware families, we compare our deep neural network models with standard image features like GIST descriptors to evaluate the performance. A joint feature measure is proposed to combine different features using error analysis to get an accurate ensemble model for improved classification performance. A new dataset called MaleX which contains around 1 million malware and benign Windows executable samples is created for large-scale malware detection and classification experiments. Experimental results are quite promising with 96% binary classification accuracy on MaleX. The proposed model is also able to generalize well on larger unseen malware samples and the results compare favorably with state-of-the-art static analysis-based malware detection algorithms.
Online continual learning for image classification studies the problem of learning to classify images from an online stream of data and tasks, where tasks may include new classes (class incremental) or data nonstationarity (domain incremental). One of the key challenges of continual learning is to avoid catastrophic forgetting (CF), i.e., forgetting old tasks in the presence of more recent tasks. Over the past few years, many methods and tricks have been introduced to address this problem, but many have not been fairly and systematically compared under a variety of realistic and practical settings. To better understand the relative advantages of various approaches and the settings where they work best, this survey aims to (1) compare state-of-the-art methods such as MIR, iCARL, and GDumb and determine which works best at different experimental settings; (2) determine if the best class incremental methods are also competitive in domain incremental setting; (3) evaluate the performance of 7 simple but effective trick such as "review" trick and nearest class mean (NCM) classifier to assess their relative impact. Regarding (1), we observe iCaRL remains competitive when the memory buffer is small; GDumb outperforms many recently proposed methods in medium-size datasets and MIR performs the best in larger-scale datasets. For (2), we note that GDumb performs quite poorly while MIR -- already competitive for (1) -- is also strongly competitive in this very different but important setting. Overall, this allows us to conclude that MIR is overall a strong and versatile method across a wide variety of settings. For (3), we find that all 7 tricks are beneficial, and when augmented with the "review" trick and NCM classifier, MIR produces performance levels that bring online continual learning much closer to its ultimate goal of matching offline training.
Surveillance scenarios are prone to several problems since they usually involve low-resolution footage, and there is no control of how far the subjects may be from the camera in the first place. This situation is suitable for the application of upsampling (super-resolution) algorithms since they may be able to recover the discriminant properties of the subjects involved. While general super-resolution approaches were proposed to enhance image quality for human-level perception, biometrics super-resolution methods seek the best "computer perception" version of the image since their focus is on improving automatic recognition performance. Convolutional neural networks and deep learning algorithms, in general, have been applied to computer vision tasks and are now state-of-the-art for several sub-domains, including image classification, restoration, and super-resolution. However, no work has evaluated the effects that the latest proposed super-resolution methods may have upon the accuracy and face verification performance in low-resolution "in-the-wild" data. This project aimed at evaluating and adapting different deep neural network architectures for the task of face super-resolution driven by face recognition performance in real-world low-resolution images. The experimental results in a real-world surveillance and attendance datasets showed that general super-resolution architectures might enhance face verification performance of deep neural networks trained on high-resolution faces. Also, since neural networks are function approximators and can be trained based on specific objective functions, the use of a customized loss function optimized for feature extraction showed promising results for recovering discriminant features in low-resolution face images.
Convolutional Neural Networks (CNN) has been extensively studied for Hyperspectral Image Classification (HSIC) more specifically, 2D and 3D CNN models have proved highly efficient in exploiting the spatial and spectral information of Hyperspectral Images. However, 2D CNN only considers the spatial information and ignores the spectral information whereas 3D CNN jointly exploits spatial-spectral information at a high computational cost. Therefore, this work proposed a lightweight CNN (3D followed by 2D-CNN) model which significantly reduces the computational cost by distributing spatial-spectral feature extraction across a lighter model alongside a preprocessing that has been carried out to improve the classification results. Five benchmark Hyperspectral datasets (i.e., SalinasA, Salinas, Indian Pines, Pavia University, Pavia Center, and Botswana) are used for experimental evaluation. The experimental results show that the proposed pipeline outperformed in terms of generalization performance, statistical significance, and computational complexity, as compared to the state-of-the-art 2D/3D CNN models except commonly used computationally expensive design choices.