Search Results for author: Kyoung Mu Lee

Found 79 papers, 36 papers with code

Attentive Fine-Grained Structured Sparsity for Image Restoration

no code implementations26 Apr 2022 Junghun Oh, Heewon Kim, Seungjun Nah, Cheeun Hong, Jonghyun Choi, Kyoung Mu Lee

Image restoration tasks have witnessed great performance improvement in recent years by developing large deep models.

Deblurring Image Restoration +1

Pay Attention to Hidden States for Video Deblurring: Ping-Pong Recurrent Neural Networks and Selective Non-Local Attention

no code implementations30 Mar 2022 JoonKyu Park, Seungjun Nah, Kyoung Mu Lee

When motion blur is strong, however, hidden states are hard to deliver proper information due to the displacement between different frames.


HandOccNet: Occlusion-Robust 3D Hand Mesh Estimation Network

no code implementations28 Mar 2022 JoonKyu Park, Yeonguk Oh, Gyeongsik Moon, Hongsuk Choi, Kyoung Mu Lee

However, we argue that occluded regions have strong correlations with hands so that they can provide highly beneficial information for complete 3D hand mesh estimation.

CVF-SID: Cyclic multi-Variate Function for Self-Supervised Image Denoising by Disentangling Noise from Image

1 code implementation24 Mar 2022 Reyhaneh Neshatavar, Mohsen Yavartanoo, Sanghyun Son, Kyoung Mu Lee

The CVF module can output multiple decomposed variables of the input and take a combination of the outputs back as an input in a cyclic manner.

Image Denoising Self-Supervised Learning

AP-BSN: Self-Supervised Denoising for Real-World Images via Asymmetric PD and Blind-Spot Network

1 code implementation22 Mar 2022 Wooseok Lee, Sanghyun Son, Kyoung Mu Lee

Extensive studies demonstrate that our method outperforms the other self-supervised and even unpaired denoising methods by a large margin, without using any additional knowledge, e. g., noise level, regarding the underlying unknown noise.


Recurrence-in-Recurrence Networks for Video Deblurring

no code implementations12 Mar 2022 JoonKyu Park, Seungjun Nah, Kyoung Mu Lee

State-of-the-art video deblurring methods often adopt recurrent neural networks to model the temporal dependency between the frames.

Deblurring Frame

C2N: Practical Generative Noise Modeling for Real-World Denoising

1 code implementation ICCV 2021 Geonwoon Jang, Wooseok Lee, Sanghyun Son, Kyoung Mu Lee

In a practical scenario, a noise generator should learn to simulate the general and complex noise distribution without using paired noisy and clean images.

Image Denoising Image Generation

Batch Normalization Tells You Which Filter is Important

no code implementations2 Dec 2021 Junghun Oh, Heewon Kim, Sungyong Baik, Cheeun Hong, Kyoung Mu Lee

The goal of filter pruning is to search for unimportant filters to remove in order to make convolutional neural networks (CNNs) efficient without sacrificing the performance in the process.

Generative Residual Attention Network for Disease Detection

no code implementations25 Oct 2021 Euyoung Kim, Soochahn Lee, Kyoung Mu Lee

Accurate identification and localization of abnormalities from radiology images serve as a critical role in computer-aided diagnosis (CAD) systems.

Meta-Learning with Task-Adaptive Loss Function for Few-Shot Learning

1 code implementation ICCV 2021 Sungyong Baik, Janghoon Choi, Heewon Kim, Dohee Cho, Jaesik Min, Kyoung Mu Lee

The problem lies in that each application and task may require different auxiliary loss function, especially when tasks are diverse and distinct.

Few-Shot Learning

Toward Real-World Super-Resolution via Adaptive Downsampling Models

no code implementations8 Sep 2021 Sanghyun Son, Jaeha Kim, Wei-Sheng Lai, Ming-Husan Yang, Kyoung Mu Lee

Most image super-resolution (SR) methods are developed on synthetic low-resolution (LR) and high-resolution (HR) image pairs that are constructed by a predetermined operation, e. g., bicubic downsampling.

Image Super-Resolution

3DIAS: 3D Shape Reconstruction with Implicit Algebraic Surfaces

1 code implementation ICCV 2021 Mohsen Yavartanoo, JaeYoung Chung, Reyhaneh Neshatavar, Kyoung Mu Lee

Our experiments demonstrate the superiorities of our method in terms of representation power compared to the state-of-the-art methods in single RGB image 3D shape reconstruction.

3D Shape Reconstruction 3D Shape Representation

NTIRE 2021 Challenge on Image Deblurring

no code implementations30 Apr 2021 Seungjun Nah, Sanghyun Son, Suyoung Lee, Radu Timofte, Kyoung Mu Lee

In this challenge report, we describe the challenge specifics and the evaluation results from the 2 competition tracks with the proposed solutions.

Deblurring Image Deblurring

NTIRE 2021 Challenge on Video Super-Resolution

no code implementations30 Apr 2021 Sanghyun Son, Suyoung Lee, Seungjun Nah, Radu Timofte, Kyoung Mu Lee

Super-Resolution (SR) is a fundamental computer vision task that aims to obtain a high-resolution clean image from the given low-resolution counterpart.

14 Frame +1

SRWarp: Generalized Image Super-Resolution under Arbitrary Transformation

1 code implementation CVPR 2021 Sanghyun Son, Kyoung Mu Lee

Deep CNNs have achieved significant successes in image processing and its applications, including single image super-resolution (SR).

Image Super-Resolution

Learning to Estimate Robust 3D Human Mesh from In-the-Wild Crowded Scenes

1 code implementation15 Apr 2021 Hongsuk Choi, Gyeongsik Moon, JoonKyu Park, Kyoung Mu Lee

Second, we propose a joint-based regressor that distinguishes a target person's feature from others.

3D Human Pose Estimation

Adaptive Dataset Sampling by Deep Policy Gradient

no code implementations1 Jan 2021 Jaerin Lee, Kyoung Mu Lee

Mini-batch SGD is a predominant optimization method in deep learning.

Image Classification

DAQ: Channel-Wise Distribution-Aware Quantization for Deep Image Super-Resolution Networks

1 code implementation21 Dec 2020 Cheeun Hong, Heewon Kim, Sungyong Baik, Junghun Oh, Kyoung Mu Lee

Quantizing deep convolutional neural networks for image super-resolution substantially reduces their computational costs.

Image Super-Resolution Quantization

Searching for Controllable Image Restoration Networks

no code implementations ICCV 2021 Heewon Kim, Sungyong Baik, Myungsub Choi, Janghoon Choi, Kyoung Mu Lee

Diverse user preferences over images have recently led to a great amount of interest in controlling the imagery effects for image restoration tasks.

Image Restoration Neural Architecture Search

NeuralAnnot: Neural Annotator for 3D Human Mesh Training Sets

3 code implementations23 Nov 2020 Gyeongsik Moon, Hongsuk Choi, Kyoung Mu Lee

Assuming no 3D pseudo-GTs are available, NeuralAnnot is weakly supervised with GT 2D/3D joint coordinates of training sets.

Accurate 3D Hand Pose Estimation for Whole-Body 3D Human Mesh Estimation

1 code implementation23 Nov 2020 Gyeongsik Moon, Hongsuk Choi, Kyoung Mu Lee

Using Pose2Pose, Hand4Whole utilizes hand MCP joint features to predict 3D wrists as MCP joints largely contribute to 3D wrist rotations in the human kinematic chain.

3D Hand Pose Estimation Pose Prediction

Beyond Static Features for Temporally Consistent 3D Human Pose and Shape from a Video

1 code implementation CVPR 2021 Hongsuk Choi, Gyeongsik Moon, Ju Yong Chang, Kyoung Mu Lee

Our TCMR significantly outperforms previous video-based methods in temporal consistency with better per-frame 3D pose and shape accuracy.

Ranked #13 on 3D Human Pose Estimation on 3DPW (using extra training data)

3D human pose and shape estimation Frame

DynaVSR: Dynamic Adaptive Blind Video Super-Resolution

1 code implementation9 Nov 2020 Suyoung Lee, Myungsub Choi, Kyoung Mu Lee

Most conventional supervised super-resolution (SR) algorithms assume that low-resolution (LR) data is obtained by downscaling high-resolution (HR) data with a fixed known kernel, but such an assumption often does not hold in real scenarios.

Frame Meta-Learning +1

Meta-Learning with Adaptive Hyperparameters

2 code implementations NeurIPS 2020 Sungyong Baik, Myungsub Choi, Janghoon Choi, Heewon Kim, Kyoung Mu Lee

Despite its popularity, several recent works question the effectiveness of MAML when test tasks are different from training tasks, thus suggesting various task-conditioned methodology to improve the initialization.

Few-Shot Learning

AIM 2020 Challenge on Video Temporal Super-Resolution

no code implementations28 Sep 2020 Sanghyun Son, Jaerin Lee, Seungjun Nah, Radu Timofte, Kyoung Mu Lee

Videos in the real-world contain various dynamics and motions that may look unnaturally discontinuous in time when the recordedframe rate is low.

Frame Super-Resolution

Domain Adaptation of Learned Features for Visual Localization

no code implementations21 Aug 2020 Sungyong Baik, Hyo Jin Kim, Tianwei Shen, Eddy Ilg, Kyoung Mu Lee, Chris Sweeney

We tackle the problem of visual localization under changing conditions, such as time of day, weather, and seasons.

Domain Adaptation Visual Localization

InterHand2.6M: A Dataset and Baseline for 3D Interacting Hand Pose Estimation from a Single RGB Image

2 code implementations ECCV 2020 Gyeongsik Moon, Shoou-I Yu, He Wen, Takaaki Shiratori, Kyoung Mu Lee

Therefore, we firstly propose (1) a large-scale dataset, InterHand2. 6M, and (2) a baseline network, InterNet, for 3D interacting hand pose estimation from a single RGB image.

3D Hand Pose Estimation

Pose2Mesh: Graph Convolutional Network for 3D Human Pose and Mesh Recovery from a 2D Human Pose

2 code implementations ECCV 2020 Hongsuk Choi, Gyeongsik Moon, Kyoung Mu Lee

Most of the recent deep learning-based 3D human pose and mesh estimation methods regress the pose and shape parameters of human mesh models, such as SMPL and MANO, from an input image.

3D Hand Pose Estimation 3D Human Pose Estimation

DeepHandMesh: A Weakly-supervised Deep Encoder-Decoder Framework for High-fidelity Hand Mesh Modeling

1 code implementation ECCV 2020 Gyeongsik Moon, Takaaki Shiratori, Kyoung Mu Lee

We design our system to be trained in an end-to-end and weakly-supervised manner; therefore, it does not require groundtruth meshes.

Visual Tracking by TridentAlign and Context Embedding

1 code implementation14 Jul 2020 Janghoon Choi, Junseok Kwon, Kyoung Mu Lee

However, extensive scale variations of the target object and distractor objects with similar categories have consistently posed challenges in visual tracking.

Frame Region Proposal +1

IntegralAction: Pose-driven Feature Integration for Robust Human Action Recognition in Videos

2 code implementations13 Jul 2020 Gyeongsik Moon, Heeseung Kwon, Kyoung Mu Lee, Minsu Cho

Most current action recognition methods heavily rely on appearance information by taking an RGB sequence of entire image regions as input.

Action Recognition Action Recognition In Videos +2

AIM 2019 Challenge on Video Temporal Super-Resolution: Methods and Results

no code implementations4 May 2020 Seungjun Nah, Sanghyun Son, Radu Timofte, Kyoung Mu Lee

Videos contain various types and strengths of motions that may look unnaturally discontinuous in time when the recorded frame rate is low.

Frame Super-Resolution

Scene-Adaptive Video Frame Interpolation via Meta-Learning

1 code implementation CVPR 2020 Myungsub Choi, Janghoon Choi, Sungyong Baik, Tae Hyun Kim, Kyoung Mu Lee

Finally, we show that our meta-learning framework can be easily employed to any video frame interpolation network and can consistently improve its performance on multiple benchmark datasets.

Frame Meta-Learning +1

Fine-Grained Neural Architecture Search

no code implementations18 Nov 2019 Heewon Kim, Seokil Hong, Bohyung Han, Heesoo Myeong, Kyoung Mu Lee

We present an elegant framework of fine-grained neural architecture search (FGNAS), which allows to employ multiple heterogeneous operations within a single layer and can even generate compositional feature maps using several different base operations.

Image Classification Image Super-Resolution +1

PoseLifter: Absolute 3D human pose lifting network from a single noisy 2D human pose

1 code implementation26 Oct 2019 Ju Yong Chang, Gyeongsik Moon, Kyoung Mu Lee

This study presents a new network (i. e., PoseLifter) that can lift a 2D human pose to an absolute 3D pose in a camera coordinate system.

3D Human Pose Estimation

Learning to Forget for Meta-Learning

1 code implementation CVPR 2020 Sungyong Baik, Seokil Hong, Kyoung Mu Lee

Model-agnostic meta-learning (MAML) tackles the problem by formulating prior knowledge as a common initialization across tasks, which is then used to quickly adapt to unseen tasks.

Few-Shot Learning

Multi-scale Aggregation R-CNN for 2D Multi-person Pose Estimation

no code implementations10 May 2019 Gyeongsik Moon, Ju Yong Chang, Kyoung Mu Lee

Multi-person pose estimation from a 2D image is challenging because it requires not only keypoint localization but also human detection.

Human Detection Multi-Person Pose Estimation

PoseFix: Model-agnostic General Human Pose Refinement Network

1 code implementation CVPR 2019 Gyeongsik Moon, Ju Yong Chang, Kyoung Mu Lee

In this paper, we propose a human pose refinement network that estimates a refined pose from a tuple of an input image and input pose.

Ranked #2 on Multi-Person Pose Estimation on COCO (Test AP metric)

Multi-Person Pose Estimation

SPNet: Deep 3D Object Classification and Retrieval using Stereographic Projection

no code implementations5 Nov 2018 Mohsen Yavartanoo, Eu Young Kim, Kyoung Mu Lee

We propose an efficient Stereographic Projection Neural Network (SPNet) for learning representations of 3D objects.

3D Object Classification Classification +1

Clustering Convolutional Kernels to Compress Deep Neural Networks

1 code implementation ECCV 2018 Sanghyun Son, Seungjun Nah, Kyoung Mu Lee

In this paper, we propose a novel method to compress CNNs by reconstructing the network from a small set of spatial convolution kernels.

General Classification Network Pruning

Task-Aware Image Downscaling

no code implementations ECCV 2018 Heewon Kim, Myungsub Choi, Bee Lim, Kyoung Mu Lee

Our framework is efficient, and it can be generalized to handle an arbitrary image resizing operation.

Colorization Super-Resolution

Deep Meta Learning for Real-Time Target-Aware Visual Tracking

no code implementations ICCV 2019 Janghoon Choi, Junseok Kwon, Kyoung Mu Lee

In this paper, we propose a novel on-line visual tracking framework based on the Siamese matching network and meta-learner network, which run at real-time speeds.

Meta-Learning Real-Time Visual Tracking

Joint Blind Motion Deblurring and Depth Estimation of Light Field

no code implementations ECCV 2018 Dongwoo Lee, Haesol Park, In Kyu Park, Kyoung Mu Lee

Removing camera motion blur from a single light field is a challenging task since it is highly ill-posed inverse problem.

Deblurring Depth Estimation

V2V-PoseNet: Voxel-to-Voxel Prediction Network for Accurate 3D Hand and Human Pose Estimation from a Single Depth Map

7 code implementations CVPR 2018 Gyeongsik Moon, Ju Yong Chang, Kyoung Mu Lee

To overcome these weaknesses, we firstly cast the 3D hand and human pose estimation problem from a single depth map into a voxel-to-voxel prediction that uses a 3D voxelized grid and estimates the per-voxel likelihood for each keypoint.

3D Hand Pose Estimation 3D Human Pose Estimation

Joint Weakly and Semi-Supervised Deep Learning for Localization and Classification of Masses in Breast Ultrasound Images

1 code implementation10 Oct 2017 Seung Yeon Shin, Soochahn Lee, Il Dong Yun, Sun Mi Kim, Kyoung Mu Lee

The results trained with only 10 strongly annotated images along with weakly annotated images were comparable to results trained from 800 strongly annotated images, with the 95% confidence interval of difference -3. 00%--5. 00%, in terms of the correct localization (CorLoc) measure, which is the ratio of images with intersection over union with ground truth higher than 0. 5.

General Classification

Look Wider to Match Image Patches with Convolutional Neural Networks

no code implementations19 Sep 2017 Haesol Park, Kyoung Mu Lee

When a human matches two images, the viewer has a natural tendency to view the wide area around the target pixel to obtain clues of right correspondence.

Stereo Matching Stereo Matching Hand

Joint Estimation of Camera Pose, Depth, Deblurring, and Super-Resolution from a Blurred Image Sequence

no code implementations ICCV 2017 Haesol Park, Kyoung Mu Lee

The conventional methods for estimating camera poses and scene structures from severely blurry or low resolution images often result in failure.

Deblurring Pose Estimation +1

Palettenet: Image recolorization with given color palette

1 code implementation The IEEE Conference on Computer Vision and Pattern Recognition (CVPR) Workshops 2017 Junho Cho, Sangdoo Yun, Kyoung Mu Lee, Jin Young Choi

PaletteNet is then designed to change the color concept of a source image so that the palette of the output image is close to the target palette.

Enhanced Deep Residual Networks for Single Image Super-Resolution

37 code implementations10 Jul 2017 Bee Lim, Sanghyun Son, Heewon Kim, Seungjun Nah, Kyoung Mu Lee

Recent research on super-resolution has progressed with the development of deep convolutional neural networks (DCNN).

Image Super-Resolution

2D-3D Pose Consistency-based Conditional Random Fields for 3D Human Pose Estimation

no code implementations13 Apr 2017 Ju Yong Chang, Kyoung Mu Lee

The unary term of the proposed CRF model is defined based on a powerful heat-map regression network, which has been proposed for 2D human pose estimation.

3D Human Pose Estimation

Real-time visual tracking by deep reinforced decision making

1 code implementation21 Feb 2017 Janghoon Choi, Junseok Kwon, Kyoung Mu Lee

In this paper, we introduce a novel real-time visual tracking algorithm based on a template selection strategy constructed by deep reinforcement learning methods.

Decision Making Frame +2

Deep Multi-scale Convolutional Neural Network for Dynamic Scene Deblurring

1 code implementation CVPR 2017 Seungjun Nah, Tae Hyun Kim, Kyoung Mu Lee

To remove these complicated motion blurs, conventional energy optimization based methods rely on simple assumptions such that blur kernel is partially uniform or locally linear.

Deblurring Image Deblurring

Occlusion-Aware Video Deblurring with a New Layered Blur Model

no code implementations29 Nov 2016 Byeongjoo Ahn, Tae Hyun Kim, Wonsik Kim, Kyoung Mu Lee

We also provide a novel analysis on the blur kernel at object boundaries, which shows the distinctive characteristics of the blur kernel that cannot be captured by conventional blur models.


Dynamic Scene Deblurring using a Locally Adaptive Linear Blur Model

no code implementations14 Mar 2016 Tae Hyun Kim, Seungjun Nah, Kyoung Mu Lee

We infer bidirectional optical flows to handle motion blurs, and also estimate Gaussian blur maps to remove optical blur from defocus in our new blur model.

Deblurring Optical Flow Estimation

Generalized Video Deblurring for Dynamic Scenes

no code implementations CVPR 2015 Tae Hyun Kim, Kyoung Mu Lee

We propose a video deblurring method to deal with general blurs inherent in dynamic scenes, contrary to other methods.

Deblurring Optical Flow Estimation

Subgraph Matching Using Compactness Prior for Robust Feature Correspondence

no code implementations CVPR 2015 Yumin Suh, Kamil Adamczewski, Kyoung Mu Lee

By constructing Markov chain on the restricted search space instead of the original solution space, our method approximates the solution effectively.

Graph Matching

MRF Optimization by Graph Approximation

no code implementations CVPR 2015 Wonsik Kim, Kyoung Mu Lee

However, not much research efforts has been done on the generation of "good" proposals, especially for non-metric energy functions.

Interval Tracker: Tracking by Interval Analysis

no code implementations CVPR 2014 Junseok Kwon, Kyoung Mu Lee

By minimizing the interval of the posterior, our method can reduce the modeling uncertainty in the posterior.

Visual Tracking

Segmentation-Free Dynamic Scene Deblurring

no code implementations CVPR 2014 Tae Hyun Kim, Kyoung Mu Lee

Thus, we propose a new energy model simultaneously estimating motion flow and the latent image based on robust total variation (TV)-L1 model.

Deblurring Motion Segmentation

Scanline Sampler without Detailed Balance: An Efficient MCMC for MRF Optimization

no code implementations CVPR 2014 Wonsik Kim, Kyoung Mu Lee

To come up with faster sampling method, we investigate two ideas: breaking detailed balance and updating multiple nodes at a time.

Minimum Uncertainty Gap for Robust Visual Tracking

no code implementations CVPR 2013 Junseok Kwon, Kyoung Mu Lee

The uncertainty of the likelihood is estimated by obtaining the gap between the lower and upper bounds of the likelihood.

Visual Tracking

Tensor-Based High-Order Semantic Relation Transfer for Semantic Scene Segmentation

no code implementations CVPR 2013 Heesoo Myeong, Kyoung Mu Lee

In this paper, we propose semantic relation transfer, a method to transfer high-order semantic relations of objects from annotated images to unlabeled images analogous to label transfer techniques where label information are transferred.

Scene Segmentation

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