Jae Shin Yoon

ORCID: 0000-0003-0181-4869
Publications
Citations
Views
---
Saved
---
About
Contact & Profiles
Research Areas
  • Advanced Vision and Imaging
  • Human Pose and Action Recognition
  • Video Surveillance and Tracking Methods
  • 3D Shape Modeling and Analysis
  • Generative Adversarial Networks and Image Synthesis
  • Face recognition and analysis
  • Advanced Neural Network Applications
  • Advanced Image and Video Retrieval Techniques
  • Autonomous Vehicle Technology and Safety
  • Human Motion and Animation
  • Image Processing and 3D Reconstruction
  • Image Enhancement Techniques
  • Visual Attention and Saliency Detection
  • Computer Graphics and Visualization Techniques
  • Image and Object Detection Techniques
  • Digital Media Forensic Detection
  • Advanced Image Processing Techniques
  • Robotics and Sensor-Based Localization
  • Educational Games and Gamification
  • Image and Video Stabilization
  • Optical measurement and interference techniques
  • Infrared Target Detection Methodologies
  • Structural Analysis and Optimization
  • Color Science and Applications
  • Advanced Materials and Mechanics

Adobe Systems (United States)
2022-2024

University of Minnesota System
2018-2022

University of Minnesota
2017-2021

Kootenay Association for Science & Technology
2017

Korea Advanced Institute of Science and Technology
2015-2017

In this paper, we propose a unified end-to-end trainable multi-task network that jointly handles lane and road marking detection recognition is guided by vanishing point under adverse weather conditions. We tackle rainy low illumination conditions, which have not been extensively studied until now due to clear challenges. For example, images taken days are subject illumination, while wet roads cause light reflection distort the appearance of markings. At night, color distortion occurs...

10.1109/iccv.2017.215 article EN 2017-10-01

We introduce the KAIST multi-spectral data set, which covers a great range of drivable regions, from urban to residential, for autonomous systems. Our set provides different perspectives world captured in coarse time slots (day and night), addition fine (sunrise, morning, afternoon, sunset, night, dawn). For all-day perception systems, we propose use spectral sensor, i.e., thermal imaging camera. Toward this goal, develop multi-sensor platform, supports co-aligned RGB/Thermal camera, RGB...

10.1109/tits.2018.2791533 article EN IEEE Transactions on Intelligent Transportation Systems 2018-02-15

We propose a novel video object segmentation algorithm based on pixel-level matching using Convolutional Neural Networks (CNN). Our network aims to distinguish the target area from background basis of similarity between two units. The proposed represents features different depth layers in order take advantage both spatial details and category-level semantic information. Furthermore, we feature compression technique that drastically reduces memory requirements while maintaining capability...

10.1109/iccv.2017.238 article EN 2017-10-01

This paper presents a new method to synthesize an image from arbitrary views and times given collection of images dynamic scene. A key challenge for the novel view synthesis arises scene reconstruction where epipolar geometry does not apply local motion contents. To address this challenge, we propose combine depth single (DSV) multi-view stereo (DMV), DSV is complete, i.e., assigned every pixel, yet view-variant in its scale, while DMV view-invariant incomplete. Our insight that although...

10.1109/cvpr42600.2020.00538 article EN 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) 2020-06-01

This paper presents a new large multiview dataset called HUMBI for human body expressions with natural clothing. The goal of is to facilitate modeling view-specific appearance and geometry gaze, face, hand, body, garment from assorted people. 107 synchronized HD cam- eras are used capture 772 distinctive subjects across gen- der, ethnicity, age, physical condition. With the mul- tiview image streams, we reconstruct high fidelity ex- pressions using 3D mesh models, which allows representing...

10.1109/cvpr42600.2020.00306 article EN 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) 2020-06-01

We present a new pose transfer method for synthesizing human animation from single image of person controlled by sequence body poses. Existing methods exhibit significant visual artifacts when applying to novel scene, resulting in temporal inconsistency and failures preserving the identity textures person. To address these limitations, we design compositional neural network that predicts silhouette, garment labels, textures. Each modular is explicitly dedicated subtask can be learned...

10.1109/cvpr46437.2021.01479 article EN 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) 2021-06-01

This paper presents a robust approach for road marking detection and recognition from images captured by an embedded camera mounted on car. Our method is designed to cope with illumination changes, shadows, harsh meteorological conditions. Furthermore, the algorithm can effectively group complex multi-symbol shapes into individual marking. For this purpose, proposed technique relies MSER features obtain candidate regions which are further merged using density-based clustering. Finally, these...

10.1109/wacv.2017.90 article EN 2017-03-01

Improvements in data-capture and face modeling techniques have enabled us to create high-fidelity realistic models. However, driving these models requires special input data, e.g., 3D meshes unwrapped textures. Also, expect clean data taken under controlled lab environments, which is very different from collected the wild. All constraints make it challenging use tracking for commodity cameras. In this paper, we propose a self-supervised domain adaptation approach enable animation of camera....

10.1109/cvpr.2019.00473 article EN 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) 2019-06-01

10.1109/cvpr52733.2024.00617 article EN 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) 2024-06-16

This paper presents a method to reconstruct complete human geometry and texture from an image of person with only partial body observed, e.g., torso. The core challenge arises the occlusion: there exists no pixel where many existing single-view reconstruction methods are not designed handle such invisible parts, leading missing data in 3D. To address this challenge, we introduce novel coarse-to-fine framework. For coarse reconstruction, explicit volumetric features learned generate 3D...

10.1109/cvpr52729.2023.00845 article EN 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) 2023-06-01

Inspired by the emergent 3D capabilities in image generators, we explore whether video generators similarly exhibit awareness. Using structure-from-motion (SfM) as a benchmark for tasks, investigate if intermediate features from OpenSora, generation model, can support camera pose estimation. We first examine native awareness routing raw outputs to SfM-prediction modules like DUSt3R. Then, impact of fine-tuning on estimation enhance Results indicate that while generator have limited inherent...

10.48550/arxiv.2501.01409 preprint EN arXiv (Cornell University) 2025-01-02

We present a lighting-aware image editing pipeline that, given portrait and text prompt, performs single relighting. Our model modifies the lighting color of both foreground background to align with provided description. The unbounded nature in creativeness allows us describe scene any sensory features including temperature, emotion, smell, time, so on. However, modeling such mapping between is extremely challenging due lack dataset where there exists no scalable data that provides large...

10.1609/aaai.v39i2.32194 article EN Proceedings of the AAAI Conference on Artificial Intelligence 2025-04-11

In this paper, we introduce a novel calibration pattern board for visible and thermal camera calibration. Our is easy to make, handy move efficient heat. Also, it preserves uniform radiance long time. Proposed method can be employed in single- multi- spectral system, also used the splitter or stereo system. As result, our shows good performance comparing with previous works, that calibrated system enough use ADAS systems.

10.1109/urai.2015.7358880 article EN 2015-10-01

Drivable region detection is challenging since various types of road, occlusion or poor illumination condition have to be considered in a outdoor environment, particularly at night. In the past decade, Many efforts been made solve these problems, however, most already existing methods are designed for visible light cameras, which inherently inefficient under low conditions. this paper, we present drivable algorithm thermal-infrared cameras order overcome aforementioned problems. The novelty...

10.1109/ivs.2016.7535507 article EN 2022 IEEE Intelligent Vehicles Symposium (IV) 2016-06-01

This paper presents a method to assign semantic label 3D reconstructed trajectory from multiview image streams. The key challenge of the labeling lies in self-occlusion and photometric inconsistency caused by object social interactions, resulting highly fragmented reconstruction with noisy labels. We address this introducing new representation called map-a probability distribution over labels per constructed set recognition across multiple views. Our conjecture is that among many views,...

10.1109/cvpr.2018.00531 article EN 2018-06-01

Appearance of dressed humans undergoes a complex geometric transformation induced not only by the static pose but also its dynamics, i.e., there exists number cloth configurations given depending on way it has moved. Such appearance modeling conditioned motion been largely neglected in existing human rendering methods, resulting physically implausible motion. A key challenge learning dynamics lies requirement prohibitively large amount observations. In this paper, we present compact...

10.1109/cvpr52688.2022.00340 article EN 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) 2022-06-01

This paper presents a new large multiview dataset called HUMBI for human body expressions with natural clothing. The goal of is to facilitate modeling view-specific appearance and geometry five primary signals including gaze, face, hand, body, garment from assorted people. 107 synchronized HD cameras are used capture 772 distinctive subjects across gender, ethnicity, age, style. With the image streams, we reconstruct using 3D mesh models, which allows representing appearance. We demonstrate...

10.1109/tpami.2021.3138762 article EN IEEE Transactions on Pattern Analysis and Machine Intelligence 2021-12-28

In this paper, we introduce a low-cost multicamera synchronization approach. Our system is to make, easy handle and convenient use. Proposed can be employed in single- multi- spectral various cameras, also used any devices which support the external trigger. As result, our shows good performance comparing with hand-eye synchronization, that synchronized images are enough use ADAS systems.

10.1109/urai.2015.7358897 article EN 2015-10-01

10.1109/cvpr52733.2024.00156 article EN 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) 2024-06-16

In this paper, we present a method of clothes retargeting; generating the potential poses and deformations given 3D clothing template model to fit onto person in single RGB image. The problem is fundamentally ill-posed as attaining ground truth data impossible, i.e., images people wearing different at exact same pose. We address challenge by utilizing large-scale synthetic generated from physical simulation, allowing us map 2D dense body pose deformation. With simulated data, propose...

10.48550/arxiv.2102.00062 preprint EN other-oa arXiv (Cornell University) 2021-01-01
Coming Soon ...