Search

Batch Whitening is a technique that accelerates and stabilizes training by transforming input features to have a zero mean (Centering) and a unit variance (Scaling), and by removing linear correlation between channels (Decorrelation). In commonly used structures, which are empirically optimized with Batch Normalization, the normalization layer appears between convolution and activation function. Following Batch Whitening studies have employed the same structure without further analysis; even Batch Whitening was analyzed on the premise that the input of a linear layer is whitened. To bridge the gap, we propose a new Convolutional Unit that is in line with the theory, and our method generally improves the performance of Batch Whitening. Moreover, we show the inefficacy of the original Convolutional Unit by investigating rank and correlation of features. As our method is employable off-the-shelf whitening modules, we use Iterative Normalization (IterNorm), the state-of-the-art whitening module, and obtain significantly improved performance on five image classification datasets: CIFAR-10, CIFAR-100, CUB-200-2011, Stanford Dogs, and ImageNet. Notably, we verify that our method improves stability and performance of whitening when using large learning rate, group size, and iteration number., Comment: ICCV 2021

We present a novel approach for oriented object detection, named TricubeNet, which localizes oriented objects using visual cues ($i.e.,$ heatmap) instead of oriented box offsets regression. We represent each object as a 2D Tricube kernel and extract bounding boxes using simple image-processing algorithms. Our approach is able to (1) obtain well-arranged boxes from visual cues, (2) solve the angle discontinuity problem, and (3) can save computational complexity due to our anchor-free modeling. To further boost the performance, we propose some effective techniques for size-invariant loss, reducing false detections, extracting rotation-invariant features, and heatmap refinement. To demonstrate the effectiveness of our TricubeNet, we experiment on various tasks for weakly-occluded oriented object detection: detection in an aerial image, densely packed object image, and text image. The extensive experimental results show that our TricubeNet is quite effective for oriented object detection. Code is available at https://github.com/qjadud1994/TricubeNet., Comment: WACV 2022, Accepted

Training of Convolutional Neural Networks (CNNs) with data with noisy labels is known to be a challenge. Based on the fact that directly providing the label to the data (Positive Learning; PL) has a risk of allowing CNNs to memorize the contaminated labels for the case of noisy data, the indirect learning approach that uses complementary labels (Negative Learning for Noisy Labels; NLNL) has proven to be highly effective in preventing overfitting to noisy data as it reduces the risk of providing faulty target. NLNL further employs a three-stage pipeline to improve convergence. As a result, filtering noisy data through the NLNL pipeline is cumbersome, increasing the training cost. In this study, we propose a novel improvement of NLNL, named Joint Negative and Positive Learning (JNPL), that unifies the filtering pipeline into a single stage. JNPL trains CNN via two losses, NL+ and PL+, which are improved upon NL and PL loss functions, respectively. We analyze the fundamental issue of NL loss function and develop new NL+ loss function producing gradient that enhances the convergence of noisy data. Furthermore, PL+ loss function is designed to enable faster convergence to expected-to-be-clean data. We show that the NL+ and PL+ train CNN simultaneously, significantly simplifying the pipeline, allowing greater ease of practical use compared to NLNL. With a simple semi-supervised training technique, our method achieves state-of-the-art accuracy for noisy data classification based on the superior filtering ability., Comment: CVPR 2021, Accepted

Weakly-supervised semantic segmentation (WSSS) using image-level labels has recently attracted much attention for reducing annotation costs. Existing WSSS methods utilize localization maps from the classification network to generate pseudo segmentation labels. However, since localization maps obtained from the classifier focus only on sparse discriminative object regions, it is difficult to generate high-quality segmentation labels. To address this issue, we introduce discriminative region suppression (DRS) module that is a simple yet effective method to expand object activation regions. DRS suppresses the attention on discriminative regions and spreads it to adjacent non-discriminative regions, generating dense localization maps. DRS requires few or no additional parameters and can be plugged into any network. Furthermore, we introduce an additional learning strategy to give a self-enhancement of localization maps, named localization map refinement learning. Benefiting from this refinement learning, localization maps are refined and enhanced by recovering some missing parts or removing noise itself. Due to its simplicity and effectiveness, our approach achieves mIoU 71.4% on the PASCAL VOC 2012 segmentation benchmark using only image-level labels. Extensive experiments demonstrate the effectiveness of our approach. The code is available at https://github.com/qjadud1994/DRS., Comment: AAAI 2021, Accepted

Self-supervised contrastive learning offers a means of learning informative features from a pool of unlabeled data. In this paper, we delve into another useful approach -- providing a way of selecting a core-set that is entirely unlabeled. In this regard, contrastive learning, one of a large number of self-supervised methods, was recently proposed and has consistently delivered the highest performance. This prompted us to choose two leading methods for contrastive learning: the simple framework for contrastive learning of visual representations (SimCLR) and the momentum contrastive (MoCo) learning framework. We calculated the cosine similarities for each example of an epoch for the entire duration of the contrastive learning process and subsequently accumulated the cosine-similarity values to obtain the coreset score. Our assumption was that an sample with low similarity would likely behave as a coreset. Compared with existing coreset selection methods with labels, our approach reduced the cost associated with human annotation. The unsupervised method implemented in this study for coreset selection obtained improved results over a randomly chosen subset, and were comparable to existing supervised coreset selection on various classification datasets (e.g., CIFAR, SVHN, and QMNIST)., Comment: 11pages

In active learning, the focus is mainly on the selection strategy of unlabeled data for enhancing the generalization capability of the next learning cycle. For this, various uncertainty measurement methods have been proposed. On the other hand, with the advent of data augmentation metrics as the regularizer on general deep learning, we notice that there can be a mutual influence between the method of unlabeled data selection and the data augmentation-based regularization techniques in active learning scenarios. Through various experiments, we confirmed that consistency-based regularization from analytical learning theory could affect the generalization capability of the classifier in combination with the existing uncertainty measurement method. By this fact, we propose a methodology to improve generalization ability, by applying data augmentation-based techniques to an active learning scenario. For the data augmentation-based regularization loss, we redefined cutout (co) and cutmix (cm) strategies as quantitative metrics and applied at both model training and unlabeled data selection steps. We have shown that the augmentation-based regularizer can lead to improved performance on the training step of active learning, while that same approach can be effectively combined with the uncertainty measurement metrics proposed so far. We used datasets such as FashionMNIST, CIFAR10, CIFAR100, and STL10 to verify the performance of the proposed active learning technique for multiple image classification tasks. Our experiments show consistent performance gains for each dataset and budget scenario.

Following considerable development in 3D scanning technologies, many studies have recently been proposed with various approaches for 3D vision tasks, including some methods that utilize 2D convolutional neural networks (CNNs). However, even though 2D CNNs have achieved high performance in many 2D vision tasks, existing works have not effectively applied them onto 3D vision tasks. In particular, segmentation has not been well studied because of the difficulty of dense prediction for each point, which requires rich feature representation. In this paper, we propose a simple and efficient architecture named point projection and back-projection network (PBP-Net), which leverages 2D CNNs for the 3D point cloud segmentation. 3 modules are introduced, each of which projects 3D point cloud onto 2D planes, extracts features using a 2D CNN backbone, and back-projects features onto the original 3D point cloud. To demonstrate effective 3D feature extraction using 2D CNN, we perform various experiments including comparison to recent methods. We analyze the proposed modules through ablation studies and perform experiments on object part segmentation (ShapeNet-Part dataset) and indoor scene semantic segmentation (S3DIS dataset). The experimental results show that proposed PBP-Net achieves comparable performance to existing state-of-the-art methods., Comment: 7 pages, accepted by IROS2020

Scene understanding is an essential technique in semantic segmentation. Although there exist several datasets that can be used for semantic segmentation, they are mainly focused on semantic image segmentation with large deep neural networks. Therefore, these networks are not useful for real time applications, especially in autonomous driving systems. In order to solve this problem, we make two contributions to semantic segmentation task. The first contribution is that we introduce the semantic video dataset, the Highway Driving dataset, which is a densely annotated benchmark for a semantic video segmentation task. The Highway Driving dataset consists of 20 video sequences having a 30Hz frame rate, and every frame is densely annotated. Secondly, we propose a baseline algorithm that utilizes a temporal correlation. Together with our attempt to analyze the temporal correlation, we expect the Highway Driving dataset to encourage research on semantic video segmentation., Comment: published on BMVC

Although well-trained deep neural networks have shown remarkable performance on numerous tasks, they rapidly forget what they have learned as soon as they begin to learn with additional data with the previous data stop being provided. In this paper, we introduce a novel algorithm, Incremental Class Learning with Attribute Sharing (ICLAS), for incremental class learning with deep neural networks. As one of its component, we also introduce a generative model, incGAN, which can generate images with increased variety compared with the training data. Under challenging environment of data deficiency, ICLAS incrementally trains classification and the generation networks. Since ICLAS trains both networks, our algorithm can perform multiple times of incremental class learning. The experiments on MNIST dataset demonstrate the advantages of our algorithm., Comment: published on ICIP

Advances in technology have led to the development of methods that can create desired visual multimedia. In particular, image generation using deep learning has been extensively studied across diverse fields. In comparison, video generation, especially on conditional inputs, remains a challenging and less explored area. To narrow this gap, we aim to train our model to produce a video corresponding to a given text description. We propose a novel training framework, Text-to-Image-to-Video Generative Adversarial Network (TiVGAN), which evolves frame-by-frame and finally produces a full-length video. In the first phase, we focus on creating a high-quality single video frame while learning the relationship between the text and an image. As the steps proceed, our model is trained gradually on more number of consecutive frames.This step-by-step learning process helps stabilize the training and enables the creation of high-resolution video based on conditional text descriptions. Qualitative and quantitative experimental results on various datasets demonstrate the effectiveness of the proposed method., Comment: IEEE Access

We propose a quadratic penalty method for continual learning of neural networks that contain batch normalization (BN) layers. The Hessian of a loss function represents the curvature of the quadratic penalty function, and a Kronecker-factored approximate curvature (K-FAC) is used widely to practically compute the Hessian of a neural network. However, the approximation is not valid if there is dependence between examples, typically caused by BN layers in deep network architectures. We extend the K-FAC method so that the inter-example relations are taken into account and the Hessian of deep neural networks can be properly approximated under practical assumptions. We also propose a method of weight merging and reparameterization to properly handle statistical parameters of BN, which plays a critical role for continual learning with BN, and a method that selects hyperparameters without source task data. Our method shows better performance than baselines in the permuted MNIST task with BN layers and in sequential learning from the ImageNet classification task to fine-grained classification tasks with ResNet-50, without any explicit or implicit use of source task data for hyperparameter selection., Comment: CVPR 2020

We propose a novel continual learning method called Residual Continual Learning (ResCL). Our method can prevent the catastrophic forgetting phenomenon in sequential learning of multiple tasks, without any source task information except the original network. ResCL reparameterizes network parameters by linearly combining each layer of the original network and a fine-tuned network; therefore, the size of the network does not increase at all. To apply the proposed method to general convolutional neural networks, the effects of batch normalization layers are also considered. By utilizing residual-learning-like reparameterization and a special weight decay loss, the trade-off between source and target performance is effectively controlled. The proposed method exhibits state-of-the-art performance in various continual learning scenarios., Comment: AAAI 2020

In this paper, we propose a Generative Translation Classification Network (GTCN) for improving visual classification accuracy in settings where classes are visually similar and data is scarce. For this purpose, we propose joint learning from a scratch to train a classifier and a generative stochastic translation network end-to-end. The translation network is used to perform on-line data augmentation across classes, whereas previous works have mostly involved domain adaptation. To help the model further benefit from this data-augmentation, we introduce an adaptive fade-in loss and a quadruplet loss. We perform experiments on multiple datasets to demonstrate the proposed method's performance in varied settings. Of particular interest, training on 40% of the dataset is enough for our model to surpass the performance of baselines trained on the full dataset. When our architecture is trained on the full dataset, we achieve comparable performance with state-of-the-art methods despite using a light-weight architecture., Comment: 14 pages, 17 figures, 13 tables

Training of deep neural networks heavily depends on the data distribution. In particular, the networks easily suffer from class imbalance. The trained networks would recognize the frequent classes better than the infrequent classes. To resolve this problem, existing approaches typically propose novel loss functions to obtain better feature embedding. In this paper, we argue that drawing a better decision boundary is as important as learning better features. Inspired by observations, we investigate how the class imbalance affects the decision boundary and deteriorates the performance. We also investigate the feature distributional discrepancy between training and test time. As a result, we propose a novel, yet simple method for class imbalanced learning. Despite its simplicity, our method shows outstanding performance. In particular, the experimental results show that we can significantly improve the network by scaling the weight vectors, even without additional training process.

Transferrable neural architecture search can be viewed as a binary optimization problem where a single optimal path should be selected among candidate paths in each edge within the repeated cell block of the directed a cyclic graph form. Recently, the field of differentiable architecture search attempts to relax the search problem continuously using a one-shot network that combines all the candidate paths in search space. However, when the one-shot network is pruned to the model in the discrete architecture space by the derivation algorithm, performance is significantly degraded to an almost random estimator. To reduce the quantization error from the heavy use of relaxation, we only sample a single edge to relax the corresponding variable and clamp variables in the other edges to zero or one. By this method, there is no performance drop after pruning the one-shot network by derivation algorithm, due to the preservation of the discrete nature of optimization variables during the search. Furthermore, the minimization of relaxation degree allows searching in a deeper network to discover better performance with remarkable search cost reduction (0.125 GPU days) compared to previous methods. By adding several regularization methods that help explore within the search space, we could obtain the network with notable performances on CIFAR-10, CIFAR-100, and ImageNet.

Training an object instance detector where only a few training object images are available is a challenging task. One solution is a cut-and-paste method that generates a training dataset by cutting object areas out of training images and pasting them onto other background images. A detector trained on a dataset generated with a cut-and-paste method suffers from the conventional domain shift problem, which stems from a discrepancy between the source domain (generated training dataset) and the target domain (real test dataset). Though state-of-the-art domain adaptation methods are able to reduce this gap, it is limited because they do not consider the difference of domain gaps of foreground and background. In this study, we present that the conventional domain gap can be divided into two sub-domain gaps for foreground and background. Then, we show that the original cut-and-paste approach suffers from a new domain gap problem, an unbalanced domain gaps, because it has two separate source domains for foreground and background, unlike the conventional domain shift problem. Then, we introduce an advanced cut-and-paste method to balance the unbalanced domain gaps by diversifying the foreground with GAN (generative adversarial network)-generated seed images and simplifying the background using image processing techniques. Experimental results show that our method is effective for balancing domain gaps and improving the accuracy of object instance detection in a cluttered indoor environment using only a few seed images. Furthermore, we show that balancing domain gaps can improve the detection accuracy of state-of-the-art domain adaptation methods., Comment: 12 pages, 7 figures

Convolutional Neural Networks (CNNs) provide excellent performance when used for image classification. The classical method of training CNNs is by labeling images in a supervised manner as in "input image belongs to this label" (Positive Learning; PL), which is a fast and accurate method if the labels are assigned correctly to all images. However, if inaccurate labels, or noisy labels, exist, training with PL will provide wrong information, thus severely degrading performance. To address this issue, we start with an indirect learning method called Negative Learning (NL), in which the CNNs are trained using a complementary label as in "input image does not belong to this complementary label." Because the chances of selecting a true label as a complementary label are low, NL decreases the risk of providing incorrect information. Furthermore, to improve convergence, we extend our method by adopting PL selectively, termed as Selective Negative Learning and Positive Learning (SelNLPL). PL is used selectively to train upon expected-to-be-clean data, whose choices become possible as NL progresses, thus resulting in superior performance of filtering out noisy data. With simple semi-supervised training technique, our method achieves state-of-the-art accuracy for noisy data classification, proving the superiority of SelNLPL's noisy data filtering ability., Comment: ICCV 2019, Accepted

We introduce Repetition-Reduction network (RRNet) for resource-constrained depth estimation, offering significantly improved efficiency in terms of computation, memory and energy consumption. The proposed method is based on repetition-reduction (RR) blocks. The RR blocks consist of the set of repeated convolutions and the residual connection layer that take place of the pointwise reduction layer with linear connection to the decoder. The RRNet help reduce memory usage and power consumption in the residual connections to the decoder layers. RRNet consumes approximately 3.84 times less energy and 3.06 times less meory and is approaximately 2.21 times faster, without increasing the demand on hardware resource relative to the baseline network (Godard et al, CVPR'17), outperforming current state-of-the-art lightweight architectures such as SqueezeNet, ShuffleNet, MobileNet and PyDNet., Comment: 9 pages, 5 figures

We propose a novel regularization algorithm to train deep neural networks, in which data at training time is severely biased. Since a neural network efficiently learns data distribution, a network is likely to learn the bias information to categorize input data. It leads to poor performance at test time, if the bias is, in fact, irrelevant to the categorization. In this paper, we formulate a regularization loss based on mutual information between feature embedding and bias. Based on the idea of minimizing this mutual information, we propose an iterative algorithm to unlearn the bias information. We employ an additional network to predict the bias distribution and train the network adversarially against the feature embedding network. At the end of learning, the bias prediction network is not able to predict the bias not because it is poorly trained, but because the feature embedding network successfully unlearns the bias information. We also demonstrate quantitative and qualitative experimental results which show that our algorithm effectively removes the bias information from feature embedding., Comment: CVPR 2019, Accepted

Convolutional networks have achieved great success in various vision tasks. This is mainly due to a considerable amount of research on network structure. In this study, instead of focusing on architectures, we focused on the convolution unit itself. The existing convolution unit has a fixed shape and is limited to observing restricted receptive fields. In earlier work, we proposed the active convolution unit (ACU), which can freely define its shape and learn by itself. In this paper, we provide a detailed analysis of the previously proposed unit and show that it is an efficient representation of a sparse weight convolution. Furthermore, we extend an ACU to a grouped ACU, which can observe multiple receptive fields in one layer. We found that the performance of a naive grouped convolution is degraded by increasing the number of groups; however, the proposed unit retains the accuracy even though the number of parameters decreases. Based on this result, we suggest a depthwise ACU, and various experiments have shown that our unit is efficient and can replace the existing convolutions.

Convolutional neural networks have achieved great success in various vision tasks; however, they incur heavy resource costs. By using deeper and wider networks, network accuracy can be improved rapidly. However, in an environment with limited resources (e.g., mobile applications), heavy networks may not be usable. This study shows that naive convolution can be deconstructed into a shift operation and pointwise convolution. To cope with various convolutions, we propose a new shift operation called active shift layer (ASL) that formulates the amount of shift as a learnable function with shift parameters. This new layer can be optimized end-to-end through backpropagation and it can provide optimal shift values. Finally, we apply this layer to a light and fast network that surpasses existing state-of-the-art networks., Comment: To be appeared in NIPS 2018

Generating a novel image by manipulating two input images is an interesting research problem in the study of generative adversarial networks (GANs). We propose a new GAN-based network that generates a fusion image with the identity of input image x and the shape of input image y. Our network can simultaneously train on more than two image datasets in an unsupervised manner. We define an identity loss LI to catch the identity of image x and a shape loss LS to get the shape of y. In addition, we propose a novel training method called Min-Patch training to focus the generator on crucial parts of an image, rather than its entirety. We show qualitative results on the VGG Youtube Pose dataset, Eye dataset (MPIIGaze and UnityEyes), and the Photo-Sketch-Cartoon dataset., Comment: CVPR 2018

Expanding the domain that deep neural network has already learned without accessing old domain data is a challenging task because deep neural networks forget previously learned information when learning new data from a new domain. In this paper, we propose a less-forgetful learning method for the domain expansion scenario. While existing domain adaptation techniques solely focused on adapting to new domains, the proposed technique focuses on working well with both old and new domains without needing to know whether the input is from the old or new domain. First, we present two naive approaches which will be problematic, then we provide a new method using two proposed properties for less-forgetful learning. Finally, we prove the effectiveness of our method through experiments on image classification tasks. All datasets used in the paper, will be released on our website for someone's follow-up study., Comment: 8 pages, accepted to AAAI 2018

Deep neural networks (DNNs) have shown the state-of-the-art level of performances in wide range of complicated tasks. In recent years, the studies have been actively conducted to analyze the black box characteristics of DNNs and to grasp the learning behaviours, tendency, and limitations of DNNs. In this paper, we investigate the limitation of DNNs in image classification task and verify it with the method inspired by cognitive psychology. Through analyzing the failure cases of ImageNet classification task, we hypothesize that the DNNs do not sufficiently learn to associate related classes of objects. To verify how DNNs understand the relatedness between object classes, we conducted experiments on the image database provided in cognitive psychology. We applied the ImageNet-trained DNNs to the database consisting of pairs of related and unrelated object images to compare the feature similarities and determine whether the pairs match each other. In the experiments, we observed that the DNNs show limited performance in determining relatedness between object classes. In addition, the DNNs present somewhat improved performance in discovering relatedness based on similarity, but they perform weaker in discovering relatedness based on association. Through these experiments, a novel analysis of learning behaviour of DNNs is provided and the limitation which needs to be overcome is suggested.

In a recent decade, ImageNet has become the most notable and powerful benchmark database in computer vision and machine learning community. As ImageNet has emerged as a representative benchmark for evaluating the performance of novel deep learning models, its evaluation tends to include only quantitative measures such as error rate, rather than qualitative analysis. Thus, there are few studies that analyze the failure cases of deep learning models in ImageNet, though there are numerous works analyzing the networks themselves and visualizing them. In this abstract, we qualitatively analyze the failure cases of ImageNet classification results from recent deep learning model, and categorize these cases according to the certain image patterns. Through this failure analysis, we believe that it can be discovered what the final challenges are in ImageNet database, which the current deep learning model is still vulnerable to., Comment: Poster presented at CVPR 2017 Scene Understanding Workshop

In recent years, deep learning has achieved great success in many computer vision applications. Convolutional neural networks (CNNs) have lately emerged as a major approach to image classification. Most research on CNNs thus far has focused on developing architectures such as the Inception and residual networks. The convolution layer is the core of the CNN, but few studies have addressed the convolution unit itself. In this paper, we introduce a convolution unit called the active convolution unit (ACU). A new convolution has no fixed shape, because of which we can define any form of convolution. Its shape can be learned through backpropagation during training. Our proposed unit has a few advantages. First, the ACU is a generalization of convolution; it can define not only all conventional convolutions, but also convolutions with fractional pixel coordinates. We can freely change the shape of the convolution, which provides greater freedom to form CNN structures. Second, the shape of the convolution is learned while training and there is no need to tune it by hand. Third, the ACU can learn better than a conventional unit, where we obtained the improvement simply by changing the conventional convolution to an ACU. We tested our proposed method on plain and residual networks, and the results showed significant improvement using our method on various datasets and architectures in comparison with the baseline., Comment: Accepted to appear in CVPR 2017

As the expressive depth of an emotional face differs with individuals or expressions, recognizing an expression using a single facial image at a moment is difficult. A relative expression of a query face compared to a reference face might alleviate this difficulty. In this paper, we propose to utilize contrastive representation that embeds a distinctive expressive factor for a discriminative purpose. The contrastive representation is calculated at the embedding layer of deep networks by comparing a given (query) image with the reference image. We attempt to utilize a generative reference image that is estimated based on the given image. Consequently, we deploy deep neural networks that embed a combination of a generative model, a contrastive model, and a discriminative model with an end-to-end training manner. In our proposed networks, we attempt to disentangle a facial expressive factor in two steps including learning of a generator network and a contrastive encoder network. We conducted extensive experiments on publicly available face expression databases (CK+, MMI, Oulu-CASIA, and in-the-wild databases) that have been widely adopted in the recent literatures. The proposed method outperforms the known state-of-the art methods in terms of the recognition accuracy.

This paper proposes a branched residual network for image classification. It is known that high-level features of deep neural network are more representative than lower-level features. By sharing the low-level features, the network can allocate more memory to high-level features. The upper layers of our proposed network are branched, so that it mimics the ensemble learning. By mimicking ensemble learning with single network, we have achieved better performance on ImageNet classification task.