Byeongkeun Kang

Cuong Manh Hoang, Byeongkeun Kang

While image segmentation is crucial in various computer vision applications, such as autonomous driving, grasping, and robot navigation, annotating all objects at the pixel-level for training is nearly impossible. Therefore, the study of unsupervised image segmentation methods is essential. In this paper, we present a pixel-level clustering framework for segmenting images into regions without using ground truth annotations. The proposed framework includes feature embedding modules with an attention mechanism, a feature statistics computing module, image reconstruction, and superpixel segmentation to achieve accurate unsupervised segmentation. Additionally, we propose a training strategy that utilizes intra-consistency within each superpixel, inter-similarity/dissimilarity between neighboring superpixels, and structural similarity between images. To avoid potential over-segmentation caused by superpixel-based losses, we also propose a post-processing method. Furthermore, we present an extension of the proposed method for unsupervised semantic segmentation. We conducted experiments on three publicly available datasets (Berkeley segmentation dataset, PASCAL VOC 2012 dataset, and COCO-Stuff dataset) to demonstrate the effectiveness of the proposed framework. The experimental results show that the proposed framework outperforms previous state-of-the-art methods.

Seongyeop Yang, Byeongkeun Kang, Yeejin Lee

Person re-identification is a problem of identifying individuals across non-overlapping cameras. Although remarkable progress has been made in the re-identification problem, it is still a challenging problem due to appearance variations of the same person as well as other people of similar appearance. Some prior works solved the issues by separating features of positive samples from features of negative ones. However, the performances of existing models considerably depend on the characteristics and statistics of the samples used for training. Thus, we propose a novel framework named sampling independent robust feature representation network (SirNet) that learns disentangled feature embedding from randomly chosen samples. A carefully designed sampling independent maximum discrepancy loss is introduced to model samples of the same person as a cluster. As a result, the proposed framework can generate additional hard negatives/positives using the learned features, which results in better discriminability from other identities. Extensive experimental results on large-scale benchmark datasets verify that the proposed model is more effective than prior state-of-the-art models.

Sejin Park, Taehyung Lee, Yeejin Lee et al.

This work addresses the task of class-incremental weakly supervised object localization (CI-WSOL). The goal is to incrementally learn object localization for novel classes using only image-level annotations while retaining the ability to localize previously learned classes. This task is important because annotating bounding boxes for every new incoming data is expensive, although object localization is crucial in various applications. To the best of our knowledge, we are the first to address this task. Thus, we first present a strong baseline method for CI-WSOL by adapting the strategies of class-incremental classifiers to mitigate catastrophic forgetting. These strategies include applying knowledge distillation, maintaining a small data set from previous tasks, and using cosine normalization. We then propose the feature drift compensation network to compensate for the effects of feature drifts on class scores and localization maps. Since updating network parameters to learn new tasks causes feature drifts, compensating for the final outputs is necessary. Finally, we evaluate our proposed method by conducting experiments on two publicly available datasets (ImageNet-100 and CUB-200). The experimental results demonstrate that the proposed method outperforms other baseline methods.

Duc Dang Trung Tran, Byeongkeun Kang, Yeejin Lee

Recently, transformer-based techniques incorporating superpoints have become prevalent in 3D instance segmentation. However, they often encounter an over-segmentation problem, especially noticeable with large objects. Additionally, unreliable mask predictions stemming from superpoint mask prediction further compound this issue. To address these challenges, we propose a novel framework called MSTA3D. It leverages multi-scale feature representation and introduces a twin-attention mechanism to effectively capture them. Furthermore, MSTA3D integrates a box query with a box regularizer, offering a complementary spatial constraint alongside semantic queries. Experimental evaluations on ScanNetV2, ScanNet200 and S3DIS datasets demonstrate that our approach surpasses state-of-the-art 3D instance segmentation methods.

Duy Tran Thanh, Yeejin Lee, Byeongkeun Kang

This work addresses the task of long-term person re-identification. Typically, person re-identification assumes that people do not change their clothes, which limits its applications to short-term scenarios. To overcome this limitation, we investigate long-term person re-identification, which considers both clothes-changing and clothes-consistent scenarios. In this paper, we propose a novel framework that effectively learns and utilizes both global and local information. The proposed framework consists of three streams: global, local body part, and head streams. The global and head streams encode identity-relevant information from an entire image and a cropped image of the head region, respectively. Both streams encode the most distinct, less distinct, and average features using the combinations of adversarial erasing, max pooling, and average pooling. The local body part stream extracts identity-related information for each body part, allowing it to be compared with the same body part from another image. Since body part annotations are not available in re-identification datasets, pseudo-labels are generated using clustering. These labels are then utilized to train a body part segmentation head in the local body part stream. The proposed framework is trained by backpropagating the weighted summation of the identity classification loss, the pair-based loss, and the pseudo body part segmentation loss. To demonstrate the effectiveness of the proposed method, we conducted experiments on three publicly available datasets (Celeb-reID, PRCC, and VC-Clothes). The experimental results demonstrate that the proposed method outperforms the previous state-of-the-art method.

Cuong Manh Hoang, Yeejin Lee, Byeongkeun Kang

This work addresses the task of generalized class discovery (GCD) in instance segmentation. The goal is to discover novel classes and obtain a model capable of segmenting instances of both known and novel categories, given labeled and unlabeled data. Since the real world contains numerous objects with long-tailed distributions, the instance distribution for each class is inherently imbalanced. To address the imbalanced distributions, we propose an instance-wise temperature assignment (ITA) method for contrastive learning and class-wise reliability criteria for pseudo-labels. The ITA method relaxes instance discrimination for samples belonging to head classes to enhance GCD. The reliability criteria are to avoid excluding most pseudo-labels for tail classes when training an instance segmentation network using pseudo-labels from GCD. Additionally, we propose dynamically adjusting the criteria to leverage diverse samples in the early stages while relying only on reliable pseudo-labels in the later stages. We also introduce an efficient soft attention module to encode object-specific representations for GCD. Finally, we evaluate our proposed method by conducting experiments on two settings: COCO$_{half}$ + LVIS and LVIS + Visual Genome. The experimental results demonstrate that the proposed method outperforms previous state-of-the-art methods.

Byeongkeun Kang, Sinhae Cha, Yeejin Lee

Weakly-supervised learning approaches have gained significant attention due to their ability to reduce the effort required for human annotations in training neural networks. This paper investigates a framework for weakly-supervised object localization, which aims to train a neural network capable of predicting both the object class and its location using only images and their image-level class labels. The proposed framework consists of a shared feature extractor, a classifier, and a localizer. The localizer predicts pixel-level class probabilities, while the classifier predicts the object class at the image level. Since image-level class labels are insufficient for training the localizer, weakly-supervised object localization methods often encounter challenges in accurately localizing the entire object region. To address this issue, the proposed method incorporates adversarial erasing and pseudo labels to improve localization accuracy. Specifically, novel losses are designed to utilize adversarially erased foreground features and adversarially erased feature maps, reducing dependence on the most discriminative region. Additionally, the proposed method employs pseudo labels to suppress activation values in the background while increasing them in the foreground. The proposed method is applied to two backbone networks (MobileNetV1 and InceptionV3) and is evaluated on three publicly available datasets (ILSVRC-2012, CUB-200-2011, and PASCAL VOC 2012). The experimental results demonstrate that the proposed method outperforms previous state-of-the-art methods across all evaluated metrics.

Hyeonseok Kim, Byeongkeun Kang, Yeejin Lee

Generalized zero-shot semantic segmentation of 3D point clouds aims to classify each point into both seen and unseen classes. A significant challenge with these models is their tendency to make biased predictions, often favoring the classes encountered during training. This problem is more pronounced in 3D applications, where the scale of the training data is typically smaller than in image-based tasks. To address this problem, we propose a novel method called E3DPC-GZSL, which reduces overconfident predictions towards seen classes without relying on separate classifiers for seen and unseen data. E3DPC-GZSL tackles the overconfidence problem by integrating an evidence-based uncertainty estimator into a classifier. This estimator is then used to adjust prediction probabilities using a dynamic calibrated stacking factor that accounts for pointwise prediction uncertainty. In addition, E3DPC-GZSL introduces a novel training strategy that improves uncertainty estimation by refining the semantic space. This is achieved by merging learnable parameters with text-derived features, thereby improving model optimization for unseen data. Extensive experiments demonstrate that the proposed approach achieves state-of-the-art performance on generalized zero-shot semantic segmentation datasets, including ScanNet v2 and S3DIS.

Cuong Manh Hoang, Yeejin Lee, Byeongkeun Kang

This work addresses the task of self-supervised learning (SSL) on a long-tailed dataset that aims to learn balanced and well-separated representations for downstream tasks such as image classification. This task is crucial because the real world contains numerous object categories, and their distributions are inherently imbalanced. Towards robust SSL on a class-imbalanced dataset, we investigate leveraging a network trained using unlabeled out-of-distribution (OOD) data that are prevalently available online. We first train a network using both in-domain (ID) and sampled OOD data by back-propagating the proposed pseudo semantic discrimination loss alongside a domain discrimination loss. The OOD data sampling and loss functions are designed to learn a balanced and well-separated embedding space. Subsequently, we further optimize the network on ID data by unsupervised contrastive learning while using the previously trained network as a guiding network. The guiding network is utilized to select positive/negative samples and to control the strengths of attractive/repulsive forces in contrastive learning. We also distil and transfer its embedding space to the training network to maintain balancedness and separability. Through experiments on four publicly available long-tailed datasets, we demonstrate that the proposed method outperforms previous state-of-the-art methods.

Giang H. Le, Anh Q. Nguyen, Byeongkeun Kang et al.

Remarkable progress has been achieved in image generation with the introduction of generative models. However, precisely controlling the content in generated images remains a challenging task due to their fundamental training objective. This paper addresses this challenge by proposing a novel image generation framework explicitly designed to incorporate desired content in output images. The framework utilizes advanced encoding techniques, integrating subnetworks called content fusion and frequency encoding modules. The frequency encoding module first captures features and structures of reference images by exclusively focusing on selected frequency components. Subsequently, the content fusion module generates a content-guiding vector that encapsulates desired content features. During the image generation process, content-guiding vectors from real images are fused with projected noise vectors. This ensures the production of generated images that not only maintain consistent content from guiding images but also exhibit diverse stylistic variations. To validate the effectiveness of the proposed framework in preserving content attributes, extensive experiments are conducted on widely used benchmark datasets, including Flickr-Faces-High Quality, Animal Faces High Quality, and Large-scale Scene Understanding datasets.

David Kim, Sinhae Cha, Byeongkeun Kang

This work addresses the task of weakly-supervised object localization. The goal is to learn object localization using only image-level class labels, which are much easier to obtain compared to bounding box annotations. This task is important because it reduces the need for labor-intensive ground-truth annotations. However, methods for object localization trained using weak supervision often suffer from limited accuracy in localization. To address this challenge and enhance localization accuracy, we propose a multiscale object localization transformer (MOLT). It comprises multiple object localization transformers that extract patch embeddings across various scales. Moreover, we introduce a deep clustering-guided refinement method that further enhances localization accuracy by utilizing separately extracted image segments. These segments are obtained by clustering pixels using convolutional neural networks. Finally, we demonstrate the effectiveness of our proposed method by conducting experiments on the publicly available ILSVRC-2012 dataset.

Yeonjun Bang, Yeejin Lee, Byeongkeun Kang

This work addresses the task of electric vehicle (EV) charging inlet detection for autonomous EV charging robots. Recently, automated EV charging systems have received huge attention to improve users' experience and to efficiently utilize charging infrastructures and parking lots. However, most related works have focused on system design, robot control, planning, and manipulation. Towards robust EV charging inlet detection, we propose a new dataset (EVCI dataset) and a novel data augmentation method that is based on image-to-image translation where typical image-to-image translation methods synthesize a new image in a different domain given an image. To the best of our knowledge, the EVCI dataset is the first EV charging inlet dataset. For the data augmentation method, we focus on being able to control synthesized images' captured environments (e.g., time, lighting) in an intuitive way. To achieve this, we first propose the environment guide vector that humans can intuitively interpret. We then propose a novel image-to-image translation network that translates a given image towards the environment described by the vector. Accordingly, it aims to synthesize a new image that has the same content as the given image while looking like captured in the provided environment by the environment guide vector. Lastly, we train a detection method using the augmented dataset. Through experiments on the EVCI dataset, we demonstrate that the proposed method outperforms the state-of-the-art methods. We also show that the proposed method is able to control synthesized images using an image and environment guide vectors.

Yoshikatsu Nakajima, Byeongkeun Kang, Hideo Saito et al.

This work addresses the task of open world semantic segmentation using RGBD sensing to discover new semantic classes over time. Although there are many types of objects in the real-word, current semantic segmentation methods make a closed world assumption and are trained only to segment a limited number of object classes. Towards a more open world approach, we propose a novel method that incrementally learns new classes for image segmentation. The proposed system first segments each RGBD frame using both color and geometric information, and then aggregates that information to build a single segmented dense 3D map of the environment. The segmented 3D map representation is a key component of our approach as it is used to discover new object classes by identifying coherent regions in the 3D map that have no semantic label. The use of coherent region in the 3D map as a primitive element, rather than traditional elements such as surfels or voxels, also significantly reduces the computational complexity and memory use of our method. It thus leads to semi-real-time performance at {10.7}Hz when incrementally updating the dense 3D map at every frame. Through experiments on the NYUDv2 dataset, we demonstrate that the proposed method is able to correctly cluster objects of both known and unseen classes. We also show the quantitative comparison with the state-of-the-art supervised methods, the processing time of each step, and the influences of each component.

Byeongkeun Kang, Subarna Tripathi, Truong Q. Nguyen

In this paper, we present a generative adversarial network framework that generates compressed images instead of synthesizing raw RGB images and compressing them separately. In the real world, most images and videos are stored and transferred in a compressed format to save storage capacity and data transfer bandwidth. However, since typical generative adversarial networks generate raw RGB images, those generated images need to be compressed by a post-processing stage to reduce the data size. Among image compression methods, JPEG has been one of the most commonly used lossy compression methods for still images. Hence, we propose a novel framework that generates JPEG compressed images using generative adversarial networks. The novel generator consists of the proposed locally connected layers, chroma subsampling layers, quantization layers, residual blocks, and convolution layers. The locally connected layer is proposed to enable block-based operations. We also discuss training strategies for the proposed architecture including the loss function and the transformation between its generator and its discriminator. The proposed method is evaluated using the publicly available CIFAR-10 dataset and LSUN bedroom dataset. The results demonstrate that the proposed method is able to generate compressed data with competitive qualities. The proposed method is a promising baseline method for joint image generation and compression using generative adversarial networks.

Byeongkeun Kang, Truong Q. Nguyen

In this work, we present a random forest framework that learns the weights, shapes, and sparsities of feature representations for real-time semantic segmentation. Typical filters (kernels) have predetermined shapes and sparsities and learn only weights. A few feature extraction methods fix weights and learn only shapes and sparsities. These predetermined constraints restrict learning and extracting optimal features. To overcome this limitation, we propose an unconstrained representation that is able to extract optimal features by learning weights, shapes, and sparsities. We, then, present the random forest framework that learns the flexible filters using an iterative optimization algorithm and segments input images using the learned representations. We demonstrate the effectiveness of the proposed method using a hand segmentation dataset for hand-object interaction and using two semantic segmentation datasets. The results show that the proposed method achieves real-time semantic segmentation using limited computational and memory resources.

Chen Du, Byeongkeun Kang, Zheng Xu et al.

De-fencing is to eliminate the captured fence on an image or a video, providing a clear view of the scene. It has been applied for many purposes including assisting photographers and improving the performance of computer vision algorithms such as object detection and recognition. However, the state-of-the-art de-fencing methods have limited performance caused by the difficulty of fence segmentation and also suffer from the motion of the camera or objects. To overcome these problems, we propose a novel method consisting of segmentation using convolutional neural networks and a fast/robust recovery algorithm. The segmentation algorithm using convolutional neural network achieves significant improvement in the accuracy of fence segmentation. The recovery algorithm using optical flow produces plausible de-fenced images and videos. The proposed method is experimented on both our diverse and complex dataset and publicly available datasets. The experimental results demonstrate that the proposed method achieves the state-of-the-art performance for both segmentation and content recovery.

Byeongkeun Kang, Yeejin Lee, Truong Q. Nguyen

In this work, we present the depth-adaptive deep neural network using a depth map for semantic segmentation. Typical deep neural networks receive inputs at the predetermined locations regardless of the distance from the camera. This fixed receptive field presents a challenge to generalize the features of objects at various distances in neural networks. Specifically, the predetermined receptive fields are too small at a short distance, and vice versa. To overcome this challenge, we develop a neural network which is able to adapt the receptive field not only for each layer but also for each neuron at the spatial location. To adjust the receptive field, we propose the depth-adaptive multiscale (DaM) convolution layer consisting of the adaptive perception neuron and the in-layer multiscale neuron. The adaptive perception neuron is to adjust the receptive field at each spatial location using the corresponding depth information. The in-layer multiscale neuron is to apply the different size of the receptive field at each feature space to learn features at multiple scales. The proposed DaM convolution is applied to two fully convolutional neural networks. We demonstrate the effectiveness of the proposed neural networks on the publicly available RGB-D dataset for semantic segmentation and the novel hand segmentation dataset for hand-object interaction. The experimental results show that the proposed method outperforms the state-of-the-art methods without any additional layers or pre/post-processing.

Subarna Tripathi, Gokce Dane, Byeongkeun Kang et al.

Deep convolutional Neural Networks (CNN) are the state-of-the-art performers for object detection task. It is well known that object detection requires more computation and memory than image classification. Thus the consolidation of a CNN-based object detection for an embedded system is more challenging. In this work, we propose LCDet, a fully-convolutional neural network for generic object detection that aims to work in embedded systems. We design and develop an end-to-end TensorFlow(TF)-based model. Additionally, we employ 8-bit quantization on the learned weights. We use face detection as a use case. Our TF-Slim based network can predict different faces of different shapes and sizes in a single forward pass. Our experimental results show that the proposed method achieves comparative accuracy comparing with state-of-the-art CNN-based face detection methods, while reducing the model size by 3x and memory-BW by ~4x comparing with one of the best real-time CNN-based object detector such as YOLO. TF 8-bit quantized model provides additional 4x memory reduction while keeping the accuracy as good as the floating point model. The proposed model thus becomes amenable for embedded implementations.

Byeongkeun Kang, Kar-Han Tan, Nan Jiang et al.

Hand segmentation for hand-object interaction is a necessary preprocessing step in many applications such as augmented reality, medical application, and human-robot interaction. However, typical methods are based on color information which is not robust to objects with skin color, skin pigment difference, and light condition variations. Thus, we propose hand segmentation method for hand-object interaction using only a depth map. It is challenging because of the small depth difference between a hand and objects during an interaction. To overcome this challenge, we propose the two-stage random decision forest (RDF) method consisting of detecting hands and segmenting hands. To validate the proposed method, we demonstrate results on the publicly available dataset of hand segmentation for hand-object interaction. The proposed method achieves high accuracy in short processing time comparing to the other state-of-the-art methods.

Byeongkeun Kang, Yeejin Lee, Truong Q. Nguyen

Real-time hand articulations tracking is important for many applications such as interacting with virtual / augmented reality devices or tablets. However, most of existing algorithms highly rely on expensive and high power-consuming GPUs to achieve real-time processing. Consequently, these systems are inappropriate for mobile and wearable devices. In this paper, we propose an efficient hand tracking system which does not require high performance GPUs. In our system, we track hand articulations by minimizing discrepancy between depth map from sensor and computer-generated hand model. We also initialize hand pose at each frame using finger detection and classification. Our contributions are: (a) propose adaptive hand model to consider different hand shapes of users without generating personalized hand model; (b) improve the highly efficient frame initialization for robust tracking and automatic initialization; (c) propose hierarchical random sampling of pixels from each depth map to improve tracking accuracy while limiting required computations. To the best of our knowledge, it is the first system that achieves both automatic hand model adjustment and real-time tracking without using GPUs.

Byeongkeun Kang, Subarna Tripathi, Truong Q. Nguyen

Sign language recognition is important for natural and convenient communication between deaf community and hearing majority. We take the highly efficient initial step of automatic fingerspelling recognition system using convolutional neural networks (CNNs) from depth maps. In this work, we consider relatively larger number of classes compared with the previous literature. We train CNNs for the classification of 31 alphabets and numbers using a subset of collected depth data from multiple subjects. While using different learning configurations, such as hyper-parameter selection with and without validation, we achieve 99.99% accuracy for observed signers and 83.58% to 85.49% accuracy for new signers. The result shows that accuracy improves as we include more data from different subjects during training. The processing time is 3 ms for the prediction of a single image. To the best of our knowledge, the system achieves the highest accuracy and speed. The trained model and dataset is available on our repository.

Yuanyuan Wu, Xiaohai He, Byeongkeun Kang et al.

This letter presents a novel approach to extract reliable dense and long-range motion trajectories of articulated human in a video sequence. Compared with existing approaches that emphasize temporal consistency of each tracked point, we also consider the spatial structure of tracked points on the articulated human. We treat points as a set of vertices, and build a triangle mesh to join them in image space. The problem of extracting long-range motion trajectories is changed to the issue of consistency of mesh evolution over time. First, self-occlusion is detected by a novel mesh-based method and an adaptive motion estimation method is proposed to initialize mesh between successive frames. Furthermore, we propose an iterative algorithm to efficiently adjust vertices of mesh for a physically plausible deformation, which can meet the local rigidity of mesh and silhouette constraints. Finally, we compare the proposed method with the state-of-the-art methods on a set of challenging sequences. Evaluations demonstrate that our method achieves favorable performance in terms of both accuracy and integrity of extracted trajectories.