Search Results for author: Junmo Kim

Found 62 papers, 21 papers with code

Test-Time Mixup Augmentation for Data and Class-Specific Uncertainty Estimation in Multi-Class Image Classification

no code implementations1 Dec 2022 Hansang Lee, Haeil Lee, Helen Hong, Junmo Kim

The experiments demonstrate that (1) the proposed data uncertainty better separates the correct and incorrect prediction than the existing uncertainty measures thanks to the mixup perturbation, and (2) the proposed class-specific uncertainty provides information on the class confusion and class similarity of the trained network for both datasets.

Image Classification

Noisy Label Classification using Label Noise Selection with Test-Time Augmentation Cross-Entropy and NoiseMix Learning

no code implementations1 Dec 2022 Hansang Lee, Haeil Lee, Helen Hong, Junmo Kim

In the classifier learning, we propose the NoiseMix method based on MixUp and BalancedMix methods by mixing the samples from the noisy and the clean label data.

Training Time Adversarial Attack Aiming the Vulnerability of Continual Learning

no code implementations29 Nov 2022 Gyojin Han, Jaehyun Choi, Hyeong Gwon Hong, Junmo Kim

Generally, regularization-based continual learning models limit access to the previous task data to imitate the real-world setting which has memory and privacy issues.

Adversarial Attack Continual Learning

AI-KD: Adversarial learning and Implicit regularization for self-Knowledge Distillation

no code implementations20 Nov 2022 Hyungmin Kim, Sungho Suh, SungHyun Baek, Daehwan Kim, Daun Jeong, Hansang Cho, Junmo Kim

Our model not only distills the deterministic and progressive knowledge which are from the pre-trained and previous epoch predictive probabilities but also transfers the knowledge of the deterministic predictive distributions using adversarial learning.

Self-Knowledge Distillation

UniCLIP: Unified Framework for Contrastive Language-Image Pre-training

no code implementations27 Sep 2022 Janghyeon Lee, Jongsuk Kim, Hyounguk Shon, Bumsoo Kim, Seung Hwan Kim, Honglak Lee, Junmo Kim

Pre-training vision-language models with contrastive objectives has shown promising results that are both scalable to large uncurated datasets and transferable to many downstream applications.

DLCFT: Deep Linear Continual Fine-Tuning for General Incremental Learning

no code implementations17 Aug 2022 Hyounguk Shon, Janghyeon Lee, Seung Hwan Kim, Junmo Kim

We show that this allows us to design a linear model where quadratic parameter regularization method is placed as the optimal continual learning policy, and at the same time enjoying the high performance of neural networks.

class-incremental learning Image Classification +1

Rethinking Efficacy of Softmax for Lightweight Non-Local Neural Networks

no code implementations27 Jul 2022 Yooshin Cho, Youngsoo Kim, Hanbyel Cho, Jaesung Ahn, Hyeong Gwon Hong, Junmo Kim

Attention maps normalized with softmax operation highly rely upon magnitude of key vectors, and performance is degenerated if the magnitude information is removed.

Stochastic Attribute Modeling for Face Super-Resolution

no code implementations16 Jul 2022 Hanbyel Cho, Yekang Lee, Jaemyung Yu, Junmo Kim

When a high-resolution (HR) image is degraded into a low-resolution (LR) image, the image loses some of the existing information.

Super-Resolution

Fix the Noise: Disentangling Source Feature for Transfer Learning of StyleGAN

1 code implementation29 Apr 2022 Dongyeun Lee, Jae Young Lee, Doyeon Kim, Jaehyun Choi, Junmo Kim

Owing to the disentangled feature space, our method can smoothly control the degree of the source features in a single model.

Transfer Learning

Spatiotemporal Augmentation on Selective Frequencies for Video Representation Learning

no code implementations8 Apr 2022 Jinhyung Kim, Taeoh Kim, Minho Shim, Dongyoon Han, Dongyoon Wee, Junmo Kim

Recent self-supervised video representation learning methods focus on maximizing the similarity between multiple augmented views from the same video and largely rely on the quality of generated views.

Action Recognition Data Augmentation +2

Projection-based Point Convolution for Efficient Point Cloud Segmentation

1 code implementation4 Feb 2022 Pyunghwan Ahn, JuYoung Yang, Eojindl Yi, Chanho Lee, Junmo Kim

Point branch consists of MLPs, while projection branch transforms point features into a 2D feature map and then apply 2D convolutions.

Point Cloud Segmentation

Global-Local Path Networks for Monocular Depth Estimation with Vertical CutDepth

3 code implementations19 Jan 2022 Doyeon Kim, Woonghyun Ka, Pyungwhan Ahn, Donggyu Joo, Sehwan Chun, Junmo Kim

Depth estimation from a single image is an important task that can be applied to various fields in computer vision, and has grown rapidly with the development of convolutional neural networks.

Monocular Depth Estimation

Self-Supervised Knowledge Transfer via Loosely Supervised Auxiliary Tasks

1 code implementation25 Oct 2021 Seungbum Hong, Jihun Yoon, Junmo Kim, Min-Kook Choi

The SSKT is independent of the network structure and dataset, and is trained differently from existing knowledge transfer methods; hence, it has an advantage in that the prior knowledge acquired from various tasks can be naturally transferred during the training process to the target task.

Transfer Learning

Cyclic Test Time Augmentation with Entropy Weight Method

no code implementations29 Sep 2021 Sewhan Chun, Jae Young Lee, Junmo Kim

The policy search method with the best level of input data dependency involves training a loss predictor network to estimate suitable transformations for each of the given input image in independent manner, resulting in instance-level transformation extraction.

Data Augmentation

Improving Generalization of Batch Whitening by Convolutional Unit Optimization

1 code implementation ICCV 2021 Yooshin Cho, Hanbyel Cho, Youngsoo Kim, Junmo Kim

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).

Image Classification

TricubeNet: 2D Kernel-Based Object Representation for Weakly-Occluded Oriented Object Detection

1 code implementation23 Apr 2021 Beomyoung Kim, Janghyeon Lee, Sihaeng Lee, Doyeon Kim, Junmo Kim

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.

object-detection Object Detection In Aerial Images +1

Joint Negative and Positive Learning for Noisy Labels

no code implementations CVPR 2021 Youngdong Kim, Juseung Yun, Hyounguk Shon, Junmo Kim

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.

Discriminative Region Suppression for Weakly-Supervised Semantic Segmentation

1 code implementation12 Mar 2021 Beomyoung Kim, Sangeun Han, Junmo Kim

However, since localization maps obtained from the classifier focus only on sparse discriminative object regions, it is difficult to generate high-quality segmentation labels.

Weakly supervised Semantic Segmentation Weakly-Supervised Semantic Segmentation

Extending Contrastive Learning to Unsupervised Coreset Selection

1 code implementation5 Mar 2021 Jeongwoo Ju, Heechul Jung, Yoonju Oh, Junmo Kim

Self-supervised contrastive learning offers a means of learning informative features from a pool of unlabeled data.

Contrastive Learning

An empirical study of a pruning mechanism

1 code implementation1 Jan 2021 Minju Jung, Hyounguk Shon, Eojindl Yi, SungHyun Baek, Junmo Kim

For the pruning and retraining phase, whether the pruned-and-retrained network benefits from the pretrained network indded is examined.

Association Network Pruning

Deep Active Learning with Augmentation-based Consistency Estimation

1 code implementation5 Nov 2020 SeulGi Hong, Heonjin Ha, Junmo Kim, Min-Kook Choi

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.

Active Learning Data Augmentation +2

Highway Driving Dataset for Semantic Video Segmentation

no code implementations2 Nov 2020 Byungju Kim, Junho Yim, Junmo Kim

Together with our attempt to analyze the temporal correlation, we expect the Highway Driving dataset to encourage research on semantic video segmentation.

Autonomous Driving Image Segmentation +4

PBP-Net: Point Projection and Back-Projection Network for 3D Point Cloud Segmentation

no code implementations2 Nov 2020 JuYoung Yang, Chanho Lee, Pyunghwan Ahn, Haeil Lee, Eojindl Yi, Junmo Kim

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.

Point Cloud Segmentation Semantic Segmentation

Collaborative Method for Incremental Learning on Classification and Generation

no code implementations29 Oct 2020 Byungju Kim, Jaeyoung Lee, KyungSu Kim, Sungjin Kim, Junmo Kim

In this paper, we introduce a novel algorithm, Incremental Class Learning with Attribute Sharing (ICLAS), for incremental class learning with deep neural networks.

Classification General Classification +1

TiVGAN: Text to Image to Video Generation with Step-by-Step Evolutionary Generator

no code implementations4 Sep 2020 Do-Yeon Kim, Donggyu Joo, Junmo Kim

Advances in technology have led to the development of methods that can create desired visual multimedia.

Image Generation Video Generation

Residual Continual Learning

no code implementations17 Feb 2020 Janghyeon Lee, Donggyu Joo, Hyeong Gwon Hong, Junmo Kim

We propose a novel continual learning method called Residual Continual Learning (ResCL).

Continual Learning

Joint Learning of Generative Translator and Classifier for Visually Similar Classes

no code implementations15 Dec 2019 ByungIn Yoo, Tristan Sylvain, Yoshua Bengio, Junmo Kim

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.

Data Augmentation Domain Adaptation +2

Adjusting Decision Boundary for Class Imbalanced Learning

1 code implementation4 Dec 2019 Byungju Kim, Junmo Kim

Inspired by observations, we investigate how the class imbalance affects the decision boundary and deteriorates the performance.

EDAS: Efficient and Differentiable Architecture Search

no code implementations3 Dec 2019 Hyeong Gwon Hong, Pyunghwan Ahn, Junmo Kim

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.

Neural Architecture Search Quantization

Cut-and-Paste Dataset Generation for Balancing Domain Gaps in Object Instance Detection

no code implementations26 Sep 2019 Woo-han Yun, Taewoo Kim, Jaeyeon Lee, Jaehong Kim, Junmo Kim

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.

Domain Adaptation Object Detection

NLNL: Negative Learning for Noisy Labels

1 code implementation ICCV 2019 Youngdong Kim, Junho Yim, Juseung Yun, Junmo Kim

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.

General Classification Image Classification

RRNet: Repetition-Reduction Network for Energy Efficient Decoder of Depth Estimation

no code implementations23 Jul 2019 Sang-Yun Oh, Hye-Jin S. Kim, Jongeun Lee, Junmo Kim

We introduce Repetition-Reduction network (RRNet) for resource-constrained depth estimation, offering significantly improved efficiency in terms of computation, memory and energy consumption.

Depth Estimation

Learning Not to Learn: Training Deep Neural Networks with Biased Data

2 code implementations CVPR 2019 Byungju Kim, Hyunwoo Kim, Kyung-Su Kim, Sungjin Kim, Junmo Kim

We propose a novel regularization algorithm to train deep neural networks, in which data at training time is severely biased.

Integrating Multiple Receptive Fields through Grouped Active Convolution

no code implementations11 Nov 2018 Yunho Jeon, Junmo Kim

Furthermore, we extend an ACU to a grouped ACU, which can observe multiple receptive fields in one layer.

Constructing Fast Network through Deconstruction of Convolution

2 code implementations NeurIPS 2018 Yunho Jeon, Junmo Kim

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.

Generating a Fusion Image: One's Identity and Another's Shape

no code implementations CVPR 2018 Donggyu Joo, Do-Yeon Kim, Junmo Kim

Generating a novel image by manipulating two input images is an interesting research problem in the study of generative adversarial networks (GANs).

Less-forgetful Learning for Domain Expansion in Deep Neural Networks

no code implementations16 Nov 2017 Heechul Jung, Jeongwoo Ju, Minju Jung, Junmo Kim

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.

Domain Adaptation Image Classification

Can Deep Neural Networks Match the Related Objects?: A Survey on ImageNet-trained Classification Models

no code implementations12 Sep 2017 Han S. Lee, Heechul Jung, Alex A. Agarwal, Junmo Kim

To verify how DNNs understand the relatedness between object classes, we conducted experiments on the image database provided in cognitive psychology.

Association General Classification +1

Why Do Deep Neural Networks Still Not Recognize These Images?: A Qualitative Analysis on Failure Cases of ImageNet Classification

no code implementations11 Sep 2017 Han S. Lee, Alex A. Agarwal, Junmo Kim

In a recent decade, ImageNet has become the most notable and powerful benchmark database in computer vision and machine learning community.

General Classification

A Gift From Knowledge Distillation: Fast Optimization, Network Minimization and Transfer Learning

1 code implementation CVPR 2017 Junho Yim, Donggyu Joo, Jihoon Bae, Junmo Kim

We introduce a novel technique for knowledge transfer, where knowledge from a pretrained deep neural network (DNN) is distilled and transferred to another DNN.

Knowledge Distillation Transfer Learning

Deep generative-contrastive networks for facial expression recognition

no code implementations21 Mar 2017 Youngsung Kim, ByungIn Yoo, Youngjun Kwak, Changkyu Choi, Junmo Kim

In this paper, we propose to utilize contrastive representation that embeds a distinctive expressive factor for a discriminative purpose.

Facial Expression Recognition

Deep Pyramidal Residual Networks

9 code implementations CVPR 2017 Dongyoon Han, Jiwhan Kim, Junmo Kim

This design, which is discussed in depth together with our new insights, has proven to be an effective means of improving generalization ability.

General Classification Image Classification

Less-forgetting Learning in Deep Neural Networks

no code implementations1 Jul 2016 Heechul Jung, Jeongwoo Ju, Minju Jung, Junmo Kim

Surprisingly, our method is very effective to forget less of the information in the source domain, and we show the effectiveness of our method using several experiments.

HARRISON: A Benchmark on HAshtag Recommendation for Real-world Images in Social Networks

1 code implementation17 May 2016 Minseok Park, Hanxiang Li, Junmo Kim

In this paper, we introduce the HARRISON dataset, a benchmark on hashtag recommendation for real world images in social networks.

Deep Saliency with Encoded Low level Distance Map and High Level Features

2 code implementations CVPR 2016 Gayoung Lee, Yu-Wing Tai, Junmo Kim

Recent advances in saliency detection have utilized deep learning to obtain high level features to detect salient regions in a scene.

Saliency Detection

Joint Fine-Tuning in Deep Neural Networks for Facial Expression Recognition

no code implementations ICCV 2015 Heechul Jung, Sihaeng Lee, Junho Yim, Sunjeong Park, Junmo Kim

Furthermore, we show that our new integration method gives more accurate results than traditional methods, such as a weighted summation and a feature concatenation method.

Facial Expression Recognition

Entropy Minimization for Convex Relaxation Approaches

no code implementations ICCV 2015 Mohamed Souiai, Martin R. Oswald, Youngwook Kee, Junmo Kim, Marc Pollefeys, Daniel Cremers

Despite their enormous success in solving hard combinatorial problems, convex relaxation approaches often suffer from the fact that the computed solutions are far from binary and that subsequent heuristic binarization may substantially degrade the quality of computed solutions.

Binarization Image Segmentation +1

Unsupervised Simultaneous Orthogonal Basis Clustering Feature Selection

no code implementations CVPR 2015 Dongyoon Han, Junmo Kim

Unlike the recent unsupervised feature selection methods, SOCFS does not explicitly use the pre-computed local structure information for data points represented as additional terms of their objective functions, but directly computes latent cluster information by the target matrix conducting orthogonal basis clustering in a single unified term of the proposed objective function.

Rotating Your Face Using Multi-Task Deep Neural Network

no code implementations CVPR 2015 Junho Yim, Heechul Jung, ByungIn Yoo, Changkyu Choi, Dusik Park, Junmo Kim

This paper proposes a new deep architecture based on a novel type of multitask learning, which can achieve superior performance in rotating to a target-pose face image from an arbitrary pose and illumination image while preserving identity.

Face Recognition

Deep Temporal Appearance-Geometry Network for Facial Expression Recognition

no code implementations5 Mar 2015 Heechul Jung, Sihaeng Lee, Sunjeong Park, Injae Lee, Chunghyun Ahn, Junmo Kim

Furthermore, one of the main contributions of this paper is that our deep network catches the facial action points automatically.

Facial Expression Recognition

Rigid Motion Segmentation using Randomized Voting

no code implementations CVPR 2014 Heechul Jung, Jeongwoo Ju, Junmo Kim

For evaluation of our algorithm, Hopkins 155 dataset, which is a representative test set for rigid motion segmentation, is adopted; it consists of two and three rigid motions.

Motion Segmentation

A Convex Relaxation of the Ambrosio--Tortorelli Elliptic Functionals for the Mumford-Shah Functional

no code implementations CVPR 2014 Youngwook Kee, Junmo Kim

In this paper, we revisit the phase-field approximation of Ambrosio and Tortorelli for the Mumford--Shah functional.

Salient Region Detection via High-Dimensional Color Transform

no code implementations CVPR 2014 Jiwhan Kim, Dongyoon Han, Yu-Wing Tai, Junmo Kim

By mapping a low dimensional RGB color to a feature vector in a high-dimensional color space, we show that we can linearly separate the salient regions from the background by finding an optimal linear combination of color coefficients in the high-dimensional color space.

Cannot find the paper you are looking for? You can Submit a new open access paper.