The problem of class imbalanced data lies in that the generalization performance of the classifier is deteriorated due to the lack of data of the minority classes. In this paper, we propose a novel minority over-sampling method to augment diversified minority samples by leveraging the rich context of the majority classes as background images. To diversify the minority samples, our key idea is to paste a foreground patch from a minority class to a background image from a majority class having affluent contexts. Our method is simple and can be easily combined with the existing long-tailed recognition methods. We empirically prove the effectiveness of the proposed oversampling method through extensive experiments and ablation studies. Without any architectural changes or complex algorithms, our method achieves state-of-the-art performance on various long-tailed classification benchmarks. Our code will be publicly available at link.
Understanding document images (e.g., invoices) has been an important research topic and has many applications in document processing automation. Through the latest advances in deep learning-based Optical Character Recognition (OCR), current Visual Document Understanding (VDU) systems have come to be designed based on OCR. Although such OCR-based approach promise reasonable performance, they suffer from critical problems induced by the OCR, e.g., (1) expensive computational costs and (2) performance degradation due to the OCR error propagation. In this paper, we propose a novel VDU model that is end-to-end trainable without underpinning OCR framework. To this end, we propose a new task and a synthetic document image generator to pre-train the model to mitigate the dependencies on large-scale real document images. Our approach achieves state-of-the-art performance on various document understanding tasks in public benchmark datasets and private industrial service datasets. Through extensive experiments and analysis, we demonstrate the effectiveness of the proposed model especially with consideration for a real-world application.
Image-mixing augmentations (e.g., Mixup or CutMix), which typically mix two images, have become de-facto training tricks for image classification. Despite their huge success on image classification, the number of images to mix has not been profoundly investigated by the previous works, only showing the naive K-image expansion leads to poor performance degradation. This paper derives a new K-image mixing augmentation based on the stick-breaking process under Dirichlet prior. We show that our method can train more robust and generalized classifiers through extensive experiments and analysis on classification accuracy, a shape of a loss landscape and adversarial robustness, than the usual two-image methods. Furthermore, we show that our probabilistic model can measure the sample-wise uncertainty and can boost the efficiency for Network Architecture Search (NAS) with 7x reduced search time.
Deep neural networks (DNNs) often rely on easy-to-learn discriminatory features, or cues, that are not necessarily essential to the problem at hand. For example, ducks in an image may be recognized based on their typical background scenery, such as lakes or streams. This phenomenon, also known as shortcut learning, is emerging as a key limitation of the current generation of machine learning models. In this work, we introduce a set of experiments to deepen our understanding of shortcut learning and its implications. We design a training setup with several shortcut cues, named WCST-ML, where each cue is equally conducive to the visual recognition problem at hand. Even under equal opportunities, we observe that (1) certain cues are preferred to others, (2) solutions biased to the easy-to-learn cues tend to converge to relatively flat minima on the loss surface, and (3) the solutions focusing on those preferred cues are far more abundant in the parameter space. We explain the abundance of certain cues via their Kolmogorov (descriptional) complexity: solutions corresponding to Kolmogorov-simple cues are abundant in the parameter space and are thus preferred by DNNs. Our studies are based on the synthetic dataset DSprites and the face dataset UTKFace. In our WCST-ML, we observe that the inborn bias of models leans toward simple cues, such as color and ethnicity. Our findings emphasize the importance of active human intervention to remove the inborn model biases that may cause negative societal impacts.
Modern image files are usually progressively transmitted and provide a preview before downloading the entire image for improved user experience to cope with a slow network connection. In this paper, with a similar goal, we propose a progressive transmission framework for deep learning models, especially to deal with the scenario where pre-trained deep learning models are transmitted from servers and executed at user devices (e.g., web browser or mobile). Our progressive transmission allows inferring approximate models in the middle of file delivery, and quickly provide an acceptable intermediate outputs. On the server-side, a deep learning model is divided and progressively transmitted to the user devices. Then, the divided pieces are progressively concatenated to construct approximate models on user devices. Experiments show that our method is computationally efficient without increasing the model size and total transmission time while preserving the model accuracy. We further demonstrate that our method can improve the user experience by providing the approximate models especially in a slow connection.
Weakly-supervised object localization (WSOL) enables finding an object using a dataset without any localization information. By simply training a classification model using only image-level annotations, the feature map of the model can be utilized as a score map for localization. In spite of many WSOL methods proposing novel strategies, there has not been any de facto standard about how to normalize the class activation map (CAM). Consequently, many WSOL methods have failed to fully exploit their own capacity because of the misuse of a normalization method. In this paper, we review many existing normalization methods and point out that they should be used according to the property of the given dataset. Additionally, we propose a new normalization method which substantially enhances the performance of any CAM-based WSOL methods. Using the proposed normalization method, we provide a comprehensive evaluation over three datasets (CUB, ImageNet and OpenImages) on three different architectures and observe significant performance gains over the conventional min-max normalization method in all the evaluated cases.
Vision Transformer (ViT) extends the application range of transformers from language processing to computer vision tasks as being an alternative architecture against the existing convolutional neural networks (CNN). Since the transformer-based architecture has been innovative for computer vision modeling, the design convention towards an effective architecture has been less studied yet. From the successful design principles of CNN, we investigate the role of the spatial dimension conversion and its effectiveness on the transformer-based architecture. We particularly attend the dimension reduction principle of CNNs; as the depth increases, a conventional CNN increases channel dimension and decreases spatial dimensions. We empirically show that such a spatial dimension reduction is beneficial to a transformer architecture as well, and propose a novel Pooling-based Vision Transformer (PiT) upon the original ViT model. We show that PiT achieves the improved model capability and generalization performance against ViT. Throughout the extensive experiments, we further show PiT outperforms the baseline on several tasks such as image classification, object detection and robustness evaluation. Source codes and ImageNet models are available at https://github.com/naver-ai/pit
ImageNet has been arguably the most popular image classification benchmark, but it is also the one with a significant level of label noise. Recent studies have shown that many samples contain multiple classes, despite being assumed to be a single-label benchmark. They have thus proposed to turn ImageNet evaluation into a multi-label task, with exhaustive multi-label annotations per image. However, they have not fixed the training set, presumably because of a formidable annotation cost. We argue that the mismatch between single-label annotations and effectively multi-label images is equally, if not more, problematic in the training setup, where random crops are applied. With the single-label annotations, a random crop of an image may contain an entirely different object from the ground truth, introducing noisy or even incorrect supervision during training. We thus re-label the ImageNet training set with multi-labels. We address the annotation cost barrier by letting a strong image classifier, trained on an extra source of data, generate the multi-labels. We utilize the pixel-wise multi-label predictions before the final pooling layer, in order to exploit the additional location-specific supervision signals. Training on the re-labeled samples results in improved model performances across the board. ResNet-50 attains the top-1 classification accuracy of 78.9% on ImageNet with our localized multi-labels, which can be further boosted to 80.2% with the CutMix regularization. We show that the models trained with localized multi-labels also outperforms the baselines on transfer learning to object detection and instance segmentation tasks, and various robustness benchmarks. The re-labeled ImageNet training set, pre-trained weights, and the source code are available at {https://github.com/naver-ai/relabel_imagenet}.
State-of-the-art video action classifiers often suffer from overfitting. They tend to be biased towards specific objects and scene cues, rather than the foreground action content, leading to sub-optimal generalization performances. Recent data augmentation strategies have been reported to address the overfitting problems in static image classifiers. Despite the effectiveness on the static image classifiers, data augmentation has rarely been studied for videos. For the first time in the field, we systematically analyze the efficacy of various data augmentation strategies on the video classification task. We then propose a powerful augmentation strategy VideoMix. VideoMix creates a new training video by inserting a video cuboid into another video. The ground truth labels are mixed proportionally to the number of voxels from each video. We show that VideoMix lets a model learn beyond the object and scene biases and extract more robust cues for action recognition. VideoMix consistently outperforms other augmentation baselines on Kinetics and the challenging Something-Something-V2 benchmarks. It also improves the weakly-supervised action localization performance on THUMOS'14. VideoMix pretrained models exhibit improved accuracies on the video detection task (AVA).
This paper addresses representational bottleneck in a network and propose a set of design principles that improves model performance significantly. We argue that a representational bottleneck may happen in a network designed by a conventional design and results in degrading the model performance. To investigate the representational bottleneck, we study the matrix rank of the features generated by ten thousand random networks. We further study the entire layer's channel configuration towards designing more accurate network architectures. Based on the investigation, we propose simple yet effective design principles to mitigate the representational bottleneck. Slight changes on baseline networks by following the principle leads to achieving remarkable performance improvements on ImageNet classification. Additionally, COCO object detection results and transfer learning results on several datasets provide other backups of the link between diminishing representational bottleneck of a network and improving performance. Code and pretrained models are available at https://github.com/clovaai/rexnet.