Discriminative self-supervised learning allows training models on any random group of internet images, and possibly recover salient information that helps differentiate between the images. Applied to ImageNet, this leads to object centric features that perform on par with supervised features on most object-centric downstream tasks. In this work, we question if using this ability, we can learn any salient and more representative information present in diverse unbounded set of images from across the globe. To do so, we train models on billions of random images without any data pre-processing or prior assumptions about what we want the model to learn. We scale our model size to dense 10 billion parameters to avoid underfitting on a large data size. We extensively study and validate our model performance on over 50 benchmarks including fairness, robustness to distribution shift, geographical diversity, fine grained recognition, image copy detection and many image classification datasets. The resulting model, not only captures well semantic information, it also captures information about artistic style and learns salient information such as geolocations and multilingual word embeddings based on visual content only. More importantly, we discover that such model is more robust, more fair, less harmful and less biased than supervised models or models trained on object centric datasets such as ImageNet.
The Image Difference Captioning (IDC) task aims to describe the visual differences between two similar images with natural language. The major challenges of this task lie in two aspects: 1) fine-grained visual differences that require learning stronger vision and language association and 2) high-cost of manual annotations that leads to limited supervised data. To address these challenges, we propose a new modeling framework following the pre-training-finetuning paradigm. Specifically, we design three self-supervised tasks and contrastive learning strategies to align visual differences and text descriptions at a fine-grained level. Moreover, we propose a data expansion strategy to utilize extra cross-task supervision information, such as data for fine-grained image classification, to alleviate the limitation of available supervised IDC data. Extensive experiments on two IDC benchmark datasets, CLEVR-Change and Birds-to-Words, demonstrate the effectiveness of the proposed modeling framework. The codes and models will be released at https://github.com/yaolinli/IDC.
Visual classification can be divided into coarse-grained and fine-grained classification. Coarse-grained classification represents categories with a large degree of dissimilarity, such as the classification of cats and dogs, while fine-grained classification represents classifications with a large degree of similarity, such as cat species, bird species, and the makes or models of vehicles. Unlike coarse-grained visual classification, fine-grained visual classification often requires professional experts to label data, which makes data more expensive. To meet this challenge, many approaches propose to automatically find the most discriminative regions and use local features to provide more precise features. These approaches only require image-level annotations, thereby reducing the cost of annotation. However, most of these methods require two- or multi-stage architectures and cannot be trained end-to-end. Therefore, we propose a novel plug-in module that can be integrated to many common backbones, including CNN-based or Transformer-based networks to provide strongly discriminative regions. The plugin module can output pixel-level feature maps and fuse filtered features to enhance fine-grained visual classification. Experimental results show that the proposed plugin module outperforms state-of-the-art approaches and significantly improves the accuracy to 92.77\% and 92.83\% on CUB200-2011 and NABirds, respectively. We have released our source code in Github https://github.com/chou141253/FGVC-PIM.git.
Recent studies have demonstrated that gradient matching-based dataset synthesis, or dataset condensation (DC), methods can achieve state-of-the-art performance when applied to data-efficient learning tasks. However, in this study, we prove that the existing DC methods can perform worse than the random selection method when task-irrelevant information forms a significant part of the training dataset. We attribute this to the lack of participation of the contrastive signals between the classes resulting from the class-wise gradient matching strategy. To address this problem, we propose Dataset Condensation with Contrastive signals (DCC) by modifying the loss function to enable the DC methods to effectively capture the differences between classes. In addition, we analyze the new loss function in terms of training dynamics by tracking the kernel velocity. Furthermore, we introduce a bi-level warm-up strategy to stabilize the optimization. Our experimental results indicate that while the existing methods are ineffective for fine-grained image classification tasks, the proposed method can successfully generate informative synthetic datasets for the same tasks. Moreover, we demonstrate that the proposed method outperforms the baselines even on benchmark datasets such as SVHN, CIFAR-10, and CIFAR-100. Finally, we demonstrate the high applicability of the proposed method by applying it to continual learning tasks.
Model pre-training is a cornerstone of modern visual recognition systems. Although fully supervised pre-training on datasets like ImageNet is still the de-facto standard, recent studies suggest that large-scale weakly supervised pre-training can outperform fully supervised approaches. This paper revisits weakly-supervised pre-training of models using hashtag supervision with modern versions of residual networks and the largest-ever dataset of images and corresponding hashtags. We study the performance of the resulting models in various transfer-learning settings including zero-shot transfer. We also compare our models with those obtained via large-scale self-supervised learning. We find our weakly-supervised models to be very competitive across all settings, and find they substantially outperform their self-supervised counterparts. We also include an investigation into whether our models learned potentially troubling associations or stereotypes. Overall, our results provide a compelling argument for the use of weakly supervised learning in the development of visual recognition systems. Our models, Supervised Weakly through hashtAGs (SWAG), are available publicly.
Hierarchical multi-granularity classification (HMC) assigns hierarchical multi-granularity labels to each object and focuses on encoding the label hierarchy, e.g., ["Albatross", "Laysan Albatross"] from coarse-to-fine levels. However, the definition of what is fine-grained is subjective, and the image quality may affect the identification. Thus, samples could be observed at any level of the hierarchy, e.g., ["Albatross"] or ["Albatross", "Laysan Albatross"], and examples discerned at coarse categories are often neglected in the conventional setting of HMC. In this paper, we study the HMC problem in which objects are labeled at any level of the hierarchy. The essential designs of the proposed method are derived from two motivations: (1) learning with objects labeled at various levels should transfer hierarchical knowledge between levels; (2) lower-level classes should inherit attributes related to upper-level superclasses. The proposed combinatorial loss maximizes the marginal probability of the observed ground truth label by aggregating information from related labels defined in the tree hierarchy. If the observed label is at the leaf level, the combinatorial loss further imposes the multi-class cross-entropy loss to increase the weight of fine-grained classification loss. Considering the hierarchical feature interaction, we propose a hierarchical residual network (HRN), in which granularity-specific features from parent levels acting as residual connections are added to features of children levels. Experiments on three commonly used datasets demonstrate the effectiveness of our approach compared to the state-of-the-art HMC approaches and fine-grained visual classification (FGVC) methods exploiting the label hierarchy.
Fine-grained Visual Classification (FGVC) aims to identify objects from subcategories. It is a very challenging task because of the subtle inter-class differences. Existing research applies large-scale convolutional neural networks or visual transformers as the feature extractor, which is extremely computationally expensive. In fact, real-world scenarios of fine-grained recognition often require a more lightweight mobile network that can be utilized offline. However, the fundamental mobile network feature extraction capability is weaker than large-scale models. In this paper, based on the lightweight MobilenetV2, we propose a Progressive Multi-Stage Interactive training method with a Recursive Mosaic Generator (RMG-PMSI). First, we propose a Recursive Mosaic Generator (RMG) that generates images with different granularities in different phases. Then, the features of different stages pass through a Multi-Stage Interaction (MSI) module, which strengthens and complements the corresponding features of different stages. Finally, using the progressive training (P), the features extracted by the model in different stages can be fully utilized and fused with each other. Experiments on three prestigious fine-grained benchmarks show that RMG-PMSI can significantly improve the performance with good robustness and transferability.
Despite great strides made on fine-grained visual classification (FGVC), current methods are still heavily reliant on fully-supervised paradigms where ample expert labels are called for. Semi-supervised learning (SSL) techniques, acquiring knowledge from unlabeled data, provide a considerable means forward and have shown great promise for coarse-grained problems. However, exiting SSL paradigms mostly assume in-distribution (i.e., category-aligned) unlabeled data, which hinders their effectiveness when re-proposed on FGVC. In this paper, we put forward a novel design specifically aimed at making out-of-distribution data work for semi-supervised FGVC, i.e., to "clue them in". We work off an important assumption that all fine-grained categories naturally follow a hierarchical structure (e.g., the phylogenetic tree of "Aves" that covers all bird species). It follows that, instead of operating on individual samples, we can instead predict sample relations within this tree structure as the optimization goal of SSL. Beyond this, we further introduced two strategies uniquely brought by these tree structures to achieve inter-sample consistency regularization and reliable pseudo-relation. Our experimental results reveal that (i) the proposed method yields good robustness against out-of-distribution data, and (ii) it can be equipped with prior arts, boosting their performance thus yielding state-of-the-art results. Code is available at https://github.com/PRIS-CV/RelMatch.
As powerful as fine-grained visual classification (FGVC) is, responding your query with a bird name of "Whip-poor-will" or "Mallard" probably does not make much sense. This however commonly accepted in the literature, underlines a fundamental question interfacing AI and human -- what constitutes transferable knowledge for human to learn from AI? This paper sets out to answer this very question using FGVC as a test bed. Specifically, we envisage a scenario where a trained FGVC model (the AI expert) functions as a knowledge provider in enabling average people (you and me) to become better domain experts ourselves, i.e. those capable in distinguishing between "Whip-poor-will" and "Mallard". Fig. 1 lays out our approach in answering this question. Assuming an AI expert trained using expert human labels, we ask (i) what is the best transferable knowledge we can extract from AI, and (ii) what is the most practical means to measure the gains in expertise given that knowledge? On the former, we propose to represent knowledge as highly discriminative visual regions that are expert-exclusive. For that, we devise a multi-stage learning framework, which starts with modelling visual attention of domain experts and novices before discriminatively distilling their differences to acquire the expert exclusive knowledge. For the latter, we simulate the evaluation process as book guide to best accommodate the learning practice of what is accustomed to humans. A comprehensive human study of 15,000 trials shows our method is able to consistently improve people of divergent bird expertise to recognise once unrecognisable birds. Interestingly, our approach also leads to improved conventional FGVC performance when the extracted knowledge defined is utilised as means to achieve discriminative localisation. Codes are available at: https://github.com/PRIS-CV/Making-a-Bird-AI-Expert-Work-for-You-and-Me
In this paper, we introduce ML-Decoder, a new attention-based classification head. ML-Decoder predicts the existence of class labels via queries, and enables better utilization of spatial data compared to global average pooling. By redesigning the decoder architecture, and using a novel group-decoding scheme, ML-Decoder is highly efficient, and can scale well to thousands of classes. Compared to using a larger backbone, ML-Decoder consistently provides a better speed-accuracy trade-off. ML-Decoder is also versatile - it can be used as a drop-in replacement for various classification heads, and generalize to unseen classes when operated with word queries. Novel query augmentations further improve its generalization ability. Using ML-Decoder, we achieve state-of-the-art results on several classification tasks: on MS-COCO multi-label, we reach 91.4% mAP; on NUS-WIDE zero-shot, we reach 31.1% ZSL mAP; and on ImageNet single-label, we reach with vanilla ResNet50 backbone a new top score of 80.7%, without extra data or distillation. Public code is available at: https://github.com/Alibaba-MIIL/ML_Decoder
Vision transformers (ViTs) have recently demonstrated state-of-the-art performance in a variety of vision tasks, replacing convolutional neural networks (CNNs). Meanwhile, since ViT has a different architecture than CNN, it may behave differently. To investigate the reliability of ViT, this paper studies the behavior and robustness of ViT. We compared the robustness of CNN and ViT by assuming various image corruptions that may appear in practical vision tasks. We confirmed that for most image transformations, ViT showed robustness comparable to CNN or more improved. However, for contrast enhancement, severe performance degradations were consistently observed in ViT. From a detailed analysis, we identified a potential problem: positional embedding in ViT's patch embedding could work improperly when the color scale changes. Here we claim the use of PreLayerNorm, a modified patch embedding structure to ensure scale-invariant behavior of ViT. ViT with PreLayerNorm showed improved robustness in various corruptions including contrast-varying environments.
The way humans attend to, process and classify a given image has the potential to vastly benefit the performance of deep learning models. Exploiting where humans are focusing can rectify models when they are deviating from essential features for correct decisions. To validate that human attention contains valuable information for decision-making processes such as fine-grained classification, we compare human attention and model explanations in discovering important features. Towards this goal, we collect human gaze data for the fine-grained classification dataset CUB and build a dataset named CUB-GHA (Gaze-based Human Attention). Furthermore, we propose the Gaze Augmentation Training (GAT) and Knowledge Fusion Network (KFN) to integrate human gaze knowledge into classification models. We implement our proposals in CUB-GHA and the recently released medical dataset CXR-Eye of chest X-ray images, which includes gaze data collected from a radiologist. Our result reveals that integrating human attention knowledge benefits classification effectively, e.g. improving the baseline by 4.38% on CXR. Hence, our work provides not only valuable insights into understanding human attention in fine-grained classification, but also contributes to future research in integrating human gaze with computer vision tasks. CUB-GHA and code are available at https://github.com/yaorong0921/CUB-GHA.
We present TWIST, a simple and theoretically explainable self-supervised representation learning method by classifying large-scale unlabeled datasets in an end-to-end way. We employ a siamese network terminated by a softmax operation to produce twin class distributions of two augmented images. Without supervision, we enforce the class distributions of different augmentations to be consistent. However, simply minimizing the divergence between augmentations will cause collapsed solutions, i.e., outputting the same class probability distribution for all images. In this case, no information about the input image is left. To solve this problem, we propose to maximize the mutual information between the input and the class predictions. Specifically, we minimize the entropy of the distribution for each sample to make the class prediction for each sample assertive and maximize the entropy of the mean distribution to make the predictions of different samples diverse. In this way, TWIST can naturally avoid the collapsed solutions without specific designs such as asymmetric network, stop-gradient operation, or momentum encoder. As a result, TWIST outperforms state-of-the-art methods on a wide range of tasks. Especially, TWIST performs surprisingly well on semi-supervised learning, achieving 61.2% top-1 accuracy with 1% ImageNet labels using a ResNet-50 as backbone, surpassing previous best results by an absolute improvement of 6.2%. Codes and pre-trained models are given on: https://github.com/bytedance/TWIST
This study aims to explore different pre-trained models offered in the Torchvision package which is available in the PyTorch library. And investigate their effectiveness on fine-grained images classification. Transfer Learning is an effective method of achieving extremely good performance with insufficient training data. In many real-world situations, people cannot collect sufficient data required to train a deep neural network model efficiently. Transfer Learning models are pre-trained on a large data set, and can bring a good performance on smaller datasets with significantly lower training time. Torchvision package offers us many models to apply the Transfer Learning on smaller datasets. Therefore, researchers may need a guideline for the selection of a good model. We investigate Torchvision pre-trained models on four different data sets: 10 Monkey Species, 225 Bird Species, Fruits 360, and Oxford 102 Flowers. These data sets have images of different resolutions, class numbers, and different achievable accuracies. We also apply their usual fully-connected layer and the Spinal fully-connected layer to investigate the effectiveness of SpinalNet. The Spinal fully-connected layer brings better performance in most situations. We apply the same augmentation for different models for the same data set for a fair comparison. This paper may help future Computer Vision researchers in choosing a proper Transfer Learning model.
Learning subtle representation about object parts plays a vital role in fine-grained visual recognition (FGVR) field. The vision transformer (ViT) achieves promising results on computer vision due to its attention mechanism. Nonetheless, with the fixed size of patches in ViT, the class token in deep layer focuses on the global receptive field and cannot generate multi-granularity features for FGVR. To capture region attention without box annotations and compensate for ViT shortcomings in FGVR, we propose a novel method named Adaptive attention multi-scale Fusion Transformer (AFTrans). The Selective Attention Collection Module (SACM) in our approach leverages attention weights in ViT and filters them adaptively to correspond with the relative importance of input patches. The multiple scales (global and local) pipeline is supervised by our weights sharing encoder and can be easily trained end-to-end. Comprehensive experiments demonstrate that AFTrans can achieve SOTA performance on three published fine-grained benchmarks: CUB-200-2011, Stanford Dogs and iNat2017.
The influential Residual Networks designed by He et al. remain the gold-standard architecture in numerous scientific publications. They typically serve as the default architecture in studies, or as baselines when new architectures are proposed. Yet there has been significant progress on best practices for training neural networks since the inception of the ResNet architecture in 2015. Novel optimization & data-augmentation have increased the effectiveness of the training recipes. In this paper, we re-evaluate the performance of the vanilla ResNet-50 when trained with a procedure that integrates such advances. We share competitive training settings and pre-trained models in the timm open-source library, with the hope that they will serve as better baselines for future work. For instance, with our more demanding training setting, a vanilla ResNet-50 reaches 80.4% top-1 accuracy at resolution 224x224 on ImageNet-val without extra data or distillation. We also report the performance achieved with popular models with our training procedure.
Traditional fine-grained image classification generally requires abundant labeled samples to deal with the low inter-class variance but high intra-class variance problem. However, in many scenarios we may have limited samples for some novel sub-categories, leading to the fine-grained few shot learning (FG-FSL) setting. To address this challenging task, we propose a novel method named foreground object transformation (FOT), which is composed of a foreground object extractor and a posture transformation generator. The former aims to remove image background, which tends to increase the difficulty of fine-grained image classification as it amplifies the intra-class variance while reduces inter-class variance. The latter transforms the posture of the foreground object to generate additional samples for the novel sub-category. As a data augmentation method, FOT can be conveniently applied to any existing few shot learning algorithm and greatly improve its performance on FG-FSL tasks. In particular, in combination with FOT, simple fine-tuning baseline methods can be competitive with the state-of-the-art methods both in inductive setting and transductive setting. Moreover, FOT can further boost the performances of latest excellent methods and bring them up to the new state-of-the-art. In addition, we also show the effectiveness of FOT on general FSL tasks.
Fine-grained classification remains a challenging task because distinguishing categories needs learning complex and local differences. Diversity in the pose, scale, and position of objects in an image makes the problem even more difficult. Although the recent Vision Transformer models achieve high performance, they need an extensive volume of input data. To encounter this problem, we made the best use of GAN-based data augmentation to generate extra dataset instances. Oxford-IIIT Pets was our dataset of choice for this experiment. It consists of 37 breeds of cats and dogs with variations in scale, poses, and lighting, which intensifies the difficulty of the classification task. Furthermore, we enhanced the performance of the recent Generative Adversarial Network (GAN), StyleGAN2-ADA model to generate more realistic images while preventing overfitting to the training set. We did this by training a customized version of MobileNetV2 to predict animal facial landmarks; then, we cropped images accordingly. Lastly, we combined the synthetic images with the original dataset and compared our proposed method with standard GANs augmentation and no augmentation with different subsets of training data. We validated our work by evaluating the accuracy of fine-grained image classification on the recent Vision Transformer (ViT) Model.
Deep neural networks have been used in various fields, but their internal behavior is not well known. In this study, we discuss two counterintuitive behaviors of convolutional neural networks (CNNs). First, we evaluated the size of the receptive field. Previous studies have attempted to increase or control the size of the receptive field. However, we observed that the size of the receptive field does not describe the classification accuracy. The size of the receptive field would be inappropriate for representing superiority in performance because it reflects only depth or kernel size and does not reflect other factors such as width or cardinality. Second, using the effective receptive field, we examined the pixels contributing to the output. Intuitively, each pixel is expected to equally contribute to the final output. However, we found that there exist pixels in a partially dead state with little contribution to the output. We reveal that the reason for this lies in the architecture of CNN and discuss solutions to reduce the phenomenon. Interestingly, for general classification tasks, the existence of dead pixels improves the training of CNNs. However, in a task that captures small perturbation, dead pixels degrade the performance. Therefore, the existence of these dead pixels should be understood and considered in practical applications of CNN.
The goal of few-shot fine-grained image classification is to recognize rarely seen fine-grained objects in the query set, given only a few samples of this class in the support set. Previous works focus on learning discriminative image features from a limited number of training samples for distinguishing various fine-grained classes, but ignore one important fact that spatial alignment of the discriminative semantic features between the query image with arbitrary changes and the support image, is also critical for computing the semantic similarity between each support-query pair. In this work, we propose an object-aware long-short-range spatial alignment approach, which is composed of a foreground object feature enhancement (FOE) module, a long-range semantic correspondence (LSC) module and a short-range spatial manipulation (SSM) module. The FOE is developed to weaken background disturbance and encourage higher foreground object response. To address the problem of long-range object feature misalignment between support-query image pairs, the LSC is proposed to learn the transferable long-range semantic correspondence by a designed feature similarity metric. Further, the SSM module is developed to refine the transformed support feature after the long-range step to align short-range misaligned features (or local details) with the query features. Extensive experiments have been conducted on four benchmark datasets, and the results show superior performance over most state-of-the-art methods under both 1-shot and 5-shot classification scenarios.