Anomaly detection can be conceived either through generative modelling of regular training data or by discriminating with respect to negative training data. These two approaches exhibit different failure modes. Consequently, hybrid algorithms present an attractive research goal. Unfortunately, dense anomaly detection requires translational equivariance and very large input resolutions. These requirements disqualify all previous hybrid approaches to the best of our knowledge. We therefore design a novel hybrid algorithm based on reinterpreting discriminative logits as a logarithm of the unnormalized joint distribution $\hat{p}(\mathbf{x},\mathbf{y})$. Our model builds on a shared convolutional representation from which we recover three dense predictions: i) the closed-set class posterior $P(\mathbf{y}|\mathbf{x})$, ii) the dataset posterior $P(\mathbf{d}_{in}|\mathbf{x})$, iii) unnormalized data likelihood $\hat{p}(\mathbf{x})$. The latter two predictions are trained both on the standard training data and on a generic negative dataset. We blend these two predictions into a hybrid anomaly score which allows dense open-set recognition on large natural images. We carefully design a custom loss for the data likelihood in order to avoid backpropagation through the untractable normalizing constant $Z(\theta)$. Experiments evaluate our contributions on standard dense anomaly detection benchmarks as well as in terms of open-mIoU - a novel metric for dense open-set performance. Our submissions achieve state-of-the-art performance despite neglectable computational overhead over the standard semantic segmentation baseline.
论文检索
输入标题、作者或关键词，从 1,645 篇学术成果中精准定位
Class Subdivision (CCS) is important for many practical applications, where the training set originally annotated for a coarse class (e.g. bird) needs to further support its sub-classes recognition (e.g. swan, crow) with only very few fine-grained labeled samples. From the perspective of causal representation learning, these sub-classes inherit the same determinative factors of the coarse class, and their difference lies only in values. Therefore, to support the challenging CCS task with minimum fine-grained labeling cost, an ideal data augmentation method should generate abundant variants by manipulating these sub-class samples at the granularity of generating factors. For this goal, traditional data augmentation methods are far from sufficient. They often perform in highly-coupled image or feature space, thus can only simulate global geometric or photometric transformations. Leveraging the recent progress of factor-disentangled generators, Unbiased Manifold Augmentation (UMA) is proposed for CCS. With a controllable StyleGAN pre-trained for a coarse class, an approximate unbiased augmentation is conducted on the factor-disentangled manifolds for each sub-class, revealing the unbiased mutual information between the target sub-class and its determinative factors. Extensive experiments have shown that in the case of small data learning (less than 1\% fine-grained samples of commonly used), our UMA can achieve 10.37\% average improvement compared with existing data augmentation methods. On challenging tasks with severe bias, the accuracy is improved by up to 16.79\%.
Attention within windows has been widely explored in vision transformers to balance the performance, computation complexity, and memory footprint. However, current models adopt a hand-crafted fixed-size window design, which restricts their capacity of modeling long-term dependencies and adapting to objects of different sizes. To address this drawback, we propose Varied-Size Window Attention (VSA) to learn adaptive window configurations from data. Specifically, based on the tokens within each default window, VSA employs a window regression module to predict the size and location of the target window, \ie, the attention area where the key and value tokens are sampled. By adopting VSA independently for each attention head, it can model long-term dependencies, capture rich context from diverse windows, and promote information exchange among overlapped windows. VSA is an easy-to-implement module that can replace the window attention in state-of-the-art representative models with minor modifications and negligible extra computational cost while improving their performance by a large margin, e.g., 1.1 % for Swin-T on ImageNet classification. In addition, the performance gain increases when using larger images for training and test. Experimental results on more downstream tasks, including object detection, instance segmentation, and semantic segmentation, further demonstrate the superiority of VSA over the vanilla window attention in dealing with objects of different sizes. The code is available at https://github.com/ViTAE-Transformer/ViTAE-VSA.
Improving Fine-Grained Visual Recognition in Low Data Regimes via Self-Boosting Attention Mechanism
PDF ↗The challenge of fine-grained visual recognition often lies in discovering the key discriminative regions. While such regions can be automatically identified from a large-scale labeled dataset, a similar method might become less effective when only a few annotations are available. In low data regimes, a network often struggles to choose the correct regions for recognition and tends to overfit spurious correlated patterns from the training data. To tackle this issue, this paper proposes the self-boosting attention mechanism, a novel method for regularizing the network to focus on the key regions shared across samples and classes. Specifically, the proposed method first generates an attention map for each training image, highlighting the discriminative part for identifying the ground-truth object category. Then the generated attention maps are used as pseudo-annotations. The network is enforced to fit them as an auxiliary task. We call this approach the self-boosting attention mechanism (SAM). We also develop a variant by using SAM to create multiple attention maps to pool convolutional maps in a style of bilinear pooling, dubbed SAM-Bilinear. Through extensive experimental studies, we show that both methods can significantly improve fine-grained visual recognition performance on low data regimes and can be incorporated into existing network architectures.
Few-shot class-incremental learning (FSCIL) aims to learn progressively about new classes with very few labeled samples, without forgetting the knowledge of already learnt classes. FSCIL suffers from two major challenges: (i) over-fitting on the new classes due to limited amount of data, (ii) catastrophically forgetting about the old classes due to unavailability of data from these classes in the incremental stages. In this work, we propose a self-supervised stochastic classifier (S3C) to counter both these challenges in FSCIL. The stochasticity of the classifier weights (or class prototypes) not only mitigates the adverse effect of absence of large number of samples of the new classes, but also the absence of samples from previously learnt classes during the incremental steps. This is complemented by the self-supervision component, which helps to learn features from the base classes which generalize well to unseen classes that are encountered in future, thus reducing catastrophic forgetting. Extensive evaluation on three benchmark datasets using multiple evaluation metrics show the effectiveness of the proposed framework. We also experiment on two additional realistic scenarios of FSCIL, namely where the number of annotated data available for each of the new classes can be different, and also where the number of base classes is much lesser, and show that the proposed S3C performs significantly better than the state-of-the-art for all these challenging scenarios.
Each year, thousands of people learn new visual categorization tasks - radiologists learn to recognize tumors, birdwatchers learn to distinguish similar species, and crowd workers learn how to annotate valuable data for applications like autonomous driving. As humans learn, their brain updates the visual features it extracts and attend to, which ultimately informs their final classification decisions. In this work, we propose a novel task of tracing the evolving classification behavior of human learners as they engage in challenging visual classification tasks. We propose models that jointly extract the visual features used by learners as well as predicting the classification functions they utilize. We collect three challenging new datasets from real human learners in order to evaluate the performance of different visual knowledge tracing methods. Our results show that our recurrent models are able to predict the classification behavior of human learners on three challenging medical image and species identification tasks.
The ability to learn new concepts continually is necessary in this ever-changing world. However, deep neural networks suffer from catastrophic forgetting when learning new categories. Many works have been proposed to alleviate this phenomenon, whereas most of them either fall into the stability-plasticity dilemma or take too much computation or storage overhead. Inspired by the gradient boosting algorithm to gradually fit the residuals between the target model and the previous ensemble model, we propose a novel two-stage learning paradigm FOSTER, empowering the model to learn new categories adaptively. Specifically, we first dynamically expand new modules to fit the residuals between the target and the output of the original model. Next, we remove redundant parameters and feature dimensions through an effective distillation strategy to maintain the single backbone model. We validate our method FOSTER on CIFAR-100 and ImageNet-100/1000 under different settings. Experimental results show that our method achieves state-of-the-art performance. Code is available at https://github.com/G-U-N/ECCV22-FOSTER.
The continual appearance of new objects in the visual world poses considerable challenges for current deep learning methods in real-world deployments. The challenge of new task learning is often exacerbated by the scarcity of data for the new categories due to rarity or cost. Here we explore the important task of Few-Shot Class-Incremental Learning (FSCIL) and its extreme data scarcity condition of one-shot. An ideal FSCIL model needs to perform well on all classes, regardless of their presentation order or paucity of data. It also needs to be robust to open-set real-world conditions and be easily adapted to the new tasks that always arise in the field. In this paper, we first reevaluate the current task setting and propose a more comprehensive and practical setting for the FSCIL task. Then, inspired by the similarity of the goals for FSCIL and modern face recognition systems, we propose our method --- Augmented Angular Loss Incremental Classification or ALICE. In ALICE, instead of the commonly used cross-entropy loss, we propose to use the angular penalty loss to obtain well-clustered features. As the obtained features not only need to be compactly clustered but also diverse enough to maintain generalization for future incremental classes, we further discuss how class augmentation, data augmentation, and data balancing affect classification performance. Experiments on benchmark datasets, including CIFAR100, miniImageNet, and CUB200, demonstrate the improved performance of ALICE over the state-of-the-art FSCIL methods.
Open set recognition (OSR) assumes unknown instances appear out of the blue at the inference time. The main challenge of OSR is that the response of models for unknowns is totally unpredictable. Furthermore, the diversity of open set makes it harder since instances have different difficulty levels. Therefore, we present a novel framework, DIfficulty-Aware Simulator (DIAS), that generates fakes with diverse difficulty levels to simulate the real world. We first investigate fakes from generative adversarial network (GAN) in the classifier’s viewpoint and observe that these are not severely challenging. This leads us to define the criteria for difficulty by regarding samples generated with GANs having moderate-difficulty. To produce hard-difficulty examples, we introduce Copycat, imitating the behavior of the classifier. Furthermore, moderate- and easy-difficulty samples are also yielded by our modified GAN and Copycat, respectively. As a result, DIAS outperforms state-of-the-art methods with both metrics of AUROC and F-score.
Recently, it is shown that deploying a proper self-supervision is a prospective way to enhance the performance of supervised learning. Yet, the benefits of self-supervision are not fully exploited as previous pretext tasks are specialized for unsupervised representation learning. To this end, we begin by presenting three desirable properties for such auxiliary tasks to assist the supervised objective. First, the tasks need to guide the model to learn rich features. Second, the transformations involved in the self-supervision should not significantly alter the training distribution. Third, the tasks are preferred to be light and generic for high applicability to prior arts. Subsequently, to show how existing pretext tasks can fulfill these and be tailored for supervised learning, we propose a simple auxiliary self-supervision task, predicting localizable rotation (\textbf{LoRot}). Our exhaustive experiments validate the merits of LoRot as a pretext task tailored for supervised learning in terms of robustness and generalization capability.
Multi-scale Vision Transformer (ViT) has emerged as a powerful backbone for computer vision tasks, while the self-attention computation in Transformer scales quadratically w.r.t. the input patch number. Thus, existing solutions commonly employ down-sampling operations (e.g., average pooling) over keys/values to dramatically reduce the computational cost. In this work, we argue that such over-aggressive down-sampling design is not invertible and inevitably causes information dropping especially for high-frequency components in objects (e.g., texture details). Motivated by the wavelet theory, we construct a new Wavelet Vision Transformer (Wave-ViT) that formulates the invertible down-sampling with wavelet transforms and self-attention learning in a unified way. This proposal enables self-attention learning with lossless down-sampling over keys/values, facilitating the pursuing of a better efficiency-vs-accuracy trade-off. Furthermore, inverse wavelet transforms are leveraged to strengthen self-attention outputs by aggregating local contexts with enlarged receptive field. We validate the superiority of Wave-ViT through extensive experiments over multiple vision tasks (e.g., image recognition, object detection and instance segmentation). Its performances surpass state-of-the-art ViT backbones with comparable FLOPs. Source code is available at \url{https://github.com/YehLi/ImageNetModel}.
Synthetic data is a scalable alternative to manual supervision, but it requires overcoming the sim-to-real domain gap. This discrepancy between virtual and real worlds is addressed by two seemingly opposed approaches: improving the realism of simulation or foregoing realism entirely via domain randomization. In this paper, we show that the recent progress in neural rendering enables a new unified approach we call Photo-realistic Neural Domain Randomization (PNDR). We propose to learn a composition of neural networks that acts as a physics-based ray tracer generating high-quality renderings from scene geometry alone. Our approach is modular, composed of different neural networks for materials, lighting, and rendering, thus enabling randomization of different key image generation components in a differentiable pipeline. Once trained, our method can be combined with other methods and used to generate photo-realistic image augmentations online and significantly more efficiently than via traditional ray-tracing. We demonstrate the usefulness of PNDR through two downstream tasks: 6D object detection and monocular depth estimation. Our experiments show that training with PNDR enables generalization to novel scenes and significantly outperforms the state of the art in terms of real-world transfer.
Overcoming Shortcut Learning in a Target Domain by Generalizing Basic Visual Factors from a Source Domain
PDF ↗Shortcut learning occurs when a deep neural network overly relies on spurious correlations in the training dataset in order to solve downstream tasks. Prior works have shown how this impairs the compositional generalization capability of deep learning models. To address this problem, we propose a novel approach to mitigate shortcut learning in uncontrolled target domains. Our approach extends the training set with an additional dataset (the source domain), which is specifically designed to facilitate learning independent representations of basic visual factors. We benchmark our idea on synthetic target domains where we explicitly control shortcut opportunities as well as real-world target domains. Furthermore, we analyze the effect of different specifications of the source domain and the network architecture on compositional generalization. Our main finding is that leveraging data from a source domain is an effective way to mitigate shortcut learning. By promoting independence across different factors of variation in the learned representations, networks can learn to consider only predictive factors and ignore potential shortcut factors during inference.
Automatic Check-Out (ACO) aims to accurately predict the presence and count of each category of products in check-out images, where a major challenge is the significant domain gap between training data (single-product exemplars) and test data (check-out images). To mitigate the gap, we propose a method, termed as PSP, to perform Prototype-based classifier learning from Single-Product exemplars. In PSP, by revealing the advantages of representing category semantics, the prototype representation of each product category is firstly obtained from single-product exemplars. Based on the prototypes, it then generates categorical classifiers with a background classifier to not only recognize fine-grained product categories but also distinguish background upon product proposals derived from check-out images. To further improve ACO accuracy, we develop discriminative re-ranking to both adjust the predicted scores of product proposals for bringing more discriminative ability in classifier learning and provide a reasonable sorting possibility by considering the fine-grained nature. Moreover, a multi-label recognition loss is also equipped for modeling co-occurrence of products in check-out images. Experiments are conducted on the large-scale RPC dataset for evaluations. Our ACO result achieves 86.69%, by 6.18% improvements over state-of-the-arts, which demonstrates the superiority of PSP. Our codes are available at https://github.com/Hao-Chen-NJUST/PSP.
Attention modules for Convolutional Neural Networks (CNNs) are an effective method to enhance performance on multiple computer-vision tasks. While existing methods appropriately model channel-, spatial- and self-attention, they primarily operate in a feedforward bottom-up manner. Consequently, the attention mechanism strongly depends on the local information of a single input feature map and does not incorporate relatively semantically-richer contextual information available at higher layers that can specify “what and where to look” in lower-level feature maps through top-down information flow. Accordingly, in this work, we propose a lightweight top-down attention module (TDAM) that iteratively generates a “visual searchlight” to perform channel and spatial modulation of its inputs and outputs more contextually-relevant feature maps at each computation step. Our experiments indicate that TDAM enhances the performance of CNNs across multiple object-recognition benchmarks and outperforms prominent attention modules while being more parameter and memory efficient. Further, TDAM-based models learn to “shift attention” by localizing individual objects or features at each computation step without any explicit supervision resulting in a 5% improvement for ResNet50 on weakly-supervised object localization. Source code and models are publicly available at: https://github.com/shantanuj/TDAM_Top_down_attention_module
Bridging Images and Videos: A Simple Learning Framework for Large Vocabulary Video Object Detection
PDF ↗Scaling object taxonomies is one of the important steps toward a robust real-world deployment of recognition systems. We have faced remarkable progress in images since the introduction of the LVIS benchmark. To continue this success in videos, a new video benchmark, TAO, was recently presented. Given the recent encouraging results from both large-vocabulary detection and tracking communities, we are interested in marrying those two advances and building a strong large vocabulary video tracker. However, supervisions in LVIS and TAO are inherently sparse or even missing, posing two new challenges for training the trackers. First, no tracking supervisions are in LVIS, which leads to inconsistent learning of detection (with LVIS and TAO) and tracking (only with TAO). Second, the detection supervisions in TAO are partial, which results in catastrophic forgetting of absent LVIS categories. To resolve these challenges, we present an effective unified learning framework that takes full advantage of all available training data to learn detection and tracking while not losing any LVIS categories to recognize. With this new learning scheme, we show that consistent improvements of various large vocabulary trackers are capable, setting strong state-of-the-art results on TAO benchmarks.
Vision Transformers (ViTs) is emerging as an alternative to convolutional neural networks (CNNs) for visual recognition. They achieve competitive results with CNNs but the lack of the typical convolutional inductive bias makes them more data-hungry than common CNNs. They are often pretrained on JFT-300M or at least ImageNet and few works study training ViTs with limited data. In this paper, we investigate how to train ViTs with limited data (e.g., 2040 images). We give theoretical analyses that our method (based on parametric instance discrimination) is superior to other methods in that it can capture both feature alignment and instance similarities. We achieve state-of-the-art results when training from scratch on 7 small datasets under various ViT backbones. We also investigate the transferring ability of small datasets and find that representations learned from small datasets can even improve large-scale ImageNet training.
We study recognizing attributes for objects in visual scenes. We consider attributes to be any phrases that describe an object’s physical and semantic properties, and its relationships with other objects. Existing work studies attribute prediction in a closed setting with a fixed set of attributes, and implements a model that uses limited context. We propose TAP, a new Transformer-based model that can utilize context and predict attributes for multiple objects in a scene in a single forward pass, and a training scheme that allows this model to learn attribute prediction from image-text datasets. Experiments on the large closed attribute benchmark VAW show that TAP outperforms the SOTA by 5.1% mAP. In addition, by utilizing pretrained text embeddings, we extend our model to OpenTAP which can recognize novel attributes not seen during training. In a large-scale setting, we further show that OpenTAP can predict a large number of seen and unseen attributes that outperforms large-scale vision-text model CLIP by a decisive margin.
Semantic novelty detection aims at discovering unknown categories in the test data. This task is particularly relevant in safety-critical applications, such as autonomous driving or healthcare, where it is crucial to recognize unknown objects at deployment time and issue a warning to the user accordingly. Despite the impressive advancements of deep learning research, existing models still need a finetuning stage on the known categories in order to recognize the unknown ones. This could be prohibitive when privacy rules limit data access, or in case of strict memory and computational constraints (e.g. edge computing). We claim that a tailored representation learning strategy may be the right solution for effective and efficient semantic novelty detection. Besides extensively testing state-of-the-art approaches for this task, we propose a novel representation learning paradigm based on relational reasoning. It focuses on learning how to measure semantic similarity rather than recognizing known categories. Our experiments show that this knowledge is directly transferable to a wide range of scenarios, and it can be exploited as a plug-and-play module to convert closed-set recognition models into reliable open-set ones.
Scene Text Recognition (STR) has many important applications in computer vision. Complex backgrounds continue to be a big challenge for STR because they interfere with text feature extraction. Many existing methods use attentional regions, bounding boxes or polygons to reduce such interference. However, the text regions located by these methods still contain much undesirable background interference. In this paper, we propose a Background-Insensitive approach BINet by explicitly leveraging the text Semantic Segmentation (SSN) to extract texts more accurately. SSN is trained on a set of existing segmentation data, whose volume is only 0.03% of STR training data. This prevents the large-scale pixel-level annotations of the STR training data. To effectively utilize the segmentation cues, we design new segmentation refinement and embedding blocks for refining text-masks and reinforcing visual features. Additionally, we propose an efficient pipeline that utilizes Synthetic Initialization (SI) for STR models trained only on real data (1.7% of STR training data), instead of on both synthetic and real data from scratch. Experiments show that the proposed method can recognize text from complex backgrounds more effectively, achieving state-of-the-art performance on several public datasets.