Jaehoon Choi

I am a Computer Science Ph.D. candidate at University of Maryland, College Park (UMD), working with Prof. Dinesh Manocha at the GAMMA Lab. I did my MS at KAIST, where I was advised by Changick Kim. I did my bachelors at the KAIST.

I am on the job market looking for research scientist, quantitative researcher and related positions. Please see my CV and contact me if you know of any opportunities that may be a good fit!

📬 kevchoi@umd.edu | jaehoonc44@gmail.com

Research

My research focuses on understanding and reconstructing the 3D world through two primary approaches: (1) generating precise 3D geometries and (2) producing photorealistic renderings of these environments.
(1) Depth Estimation and Completion: SelfDeco (ICRA'21), DnD (ICCV'21), SelfTune (ICRA'22)
(2) Neural Reconstruction and Rendering: TMO (CVPR'23), LTM (CVPR'24), MeshGS (ACCV'24), Mode-GS (Preprint)
(3) Neural Data Generation and Domain Adaptation: UAV-Sim (ICRA'24), SEGDA (ICCV'19), STUDA (ICCV'19)
(4) 360 Imaging: EDM (Submitted), IM360 (Submitted)

Publications

IM360: Textured Mesh Reconstruction for Large-scale Indoor Mapping with 360° Cameras
Dongki Jung*, Jaehoon Choi*, Yonghan Lee, Dinesh Manocha (* These two authors contributed equally)
[Project]

We propose a complete pipeline for indoor mapping using omnidirectional images, consisting of three key stages: (1) Spherical SfM, (2) Neural Surface Reconstruction, and (3) Texture Optimization.

EDM: Equirectangular Projection-Oriented Dense Kernelized Feature Matching
Dongki Jung, Jaehoon Choi, Yonghan Lee, Somi Jeong, Taejae Lee, Dinesh Manocha, Suyong Yeon
[Project]

We propose the first learning-based dense matching algorithm for omnidirectional images.

Mode-GS: Monocular Depth Guided Anchored 3D Gaussian Splatting for Robust Ground-View Scene Rendering
Yonghan Lee, Jaehoon Choi, Dongki Jung, Jaeseong Yun, Soohyun Ryu, Dinesh Manocha, and Suyong Yeon
[arXiv]

We propose a novel 3D Gaussian splatting algorithm that integrates monocular depth network with anchored Gaussian splatting, enabling robust rendering performance on sparse-view datasets.

TK-Planes: Tiered K-Planes with High Dimensional Feature Vectors for Dynamic UAV-based Scenes
Christopher Maxey, Jaehoon Choi, Yonghan Lee, Hyungtae Lee, Dinesh Manocha, and Heesung Kwon
[arXiv]

We propose an extension of K-Planes Neural Radiance Field (NeRF), wherein our algorithm stores a set of tiered feature vectors.

MeshGS: Adaptive Mesh-Aligned Gaussian Splatting for High-Quality Rendering
Jaehoon Choi, Yonghan Lee, Hyungtae Lee, Heesung Kwon, and Dinesh Manocha
ACCV, 2024
[arXiv]

We propose a novel approach that integrates mesh representation with 3D Gaussian splats to perform high-quality rendering of reconstructed real-world scenes.

LTM: Lightweight Textured Mesh Extraction and Refinement of Large Unbounded Scenes for Efficient Storage and Real-time Rendering
Jaehoon Choi, Rajvi Shah, Qinbo Li, Yipeng Wang, Ayush Saraf, Changil Kim, Jia-Bin Huang, Dinesh Manocha, Suhib Alsisan, and Johannes Kopf
CVPR, 2024
[arXiv] [Project]

We present a practical method for reconstructing and optimizing textured meshes of large, unbounded real-world scenes that offer high visual and geometric fidelity.

UAV-Sim: NeRF-based Synthetic Data Generation for UAV-based Perception
Christopher Maxey*, Jaehoon Choi*, Hyungtae Lee, Dinesh Manocha, Heesung Kwon (* These two authors contributed equally)
ICRA, 2024
[arXiv] [Video]

We present a synthetic data generation pipeline based on NeRF for UAV-based perception

TMO: Textured Mesh Acquisition of Objects with a Mobile Device by using Differentiable Rendering
Jaehoon Choi, Dongki Jung, Taejae Lee, Sangwook Kim, Youngdong Jung, Dinesh Manocha, Donghwan Lee
CVPR, 2023
[arXiv] [Project]

We present a new pipeline for acquiring a textured mesh in the wild with a single smartphone.

SelfTune: Metrically Scaled Monocular Depth Estimation through Self-Supervised Learning
Jaehoon Choi*, Dongki Jung*, Yonghan Lee, Deokhwa Kim, Dinesh Manocha, Donghwan Lee (* These two authors contributed equally)
ICRA, 2022
[arXiv]

We have developed a fine-tuning method for metrically accurate depth estimation in a self-supervised way.

DnD: Dense Depth Estimation in Crowded Dynamic Indoor Scenes
Dongki Jung*, Jaehoon Choi*, Yonghan Lee, Deokhwa Kim, Changick Kim, Dinesh Manocha, Donghwan Lee (* These two authors contributed equally)
ICCV, 2021
[arXiv]

We present a novel approach for estimating depth from a monocular camera as it moves through complex and crowded indoor environments.

Just a Few Points are All You Need for Multi-view Stereo: A Novel Semi-supervised Learning Method for Multi-view Stereo
Taekyung Kim, Jaehoon Choi, Seokeon Choi, Dongki Jung, Changick Kim
ICCV, 2021
[arXiv]

We first introduce a novel semi-supervised multi-view stereo framework.

SelfDeco: Self-Supervised Monocular Depth Completion in Challenging Indoor Environments
Jaehoon Choi, Dongki Jung, Yonghan Lee, Deokhwa Kim, Dinesh Manocha, Donghwan Lee
ICRA, 2021
[arXiv]

We present a novel algorithm for self-supervised monocular depth completion in challenging indoor environments.

SAFENet: Self-Supervised Monocular Depth Estimation with Semantic-Aware Feature Extraction
Jaehoon Choi*, Dongki Jung*, Donghwan Lee, Changick Kim (* These two authors contributed equally)
NeurIPS Workshop on Machine Learning for Autonomous Driving, 2020
[arXiv]

We propose SAFENet that is designed to leverage semantic information to overcome the limitations of the photometric loss.

Arbitrary Style Transfer Using Graph Instance Normalization
Dongki Jung, Seunghan Yang, Jaehoon Choi, Changick Kim
ICIP, 2020
[arXiv]

We present a novel learnable normalization technique for style transfer using graph convolutional networks

Self-Ensembling with GAN-based Data Augmentation for Domain Adaptation in Semantic Segmentation
Jaehoon Choi, Taekyung Kim, Changick Kim
ICCV, 2019
[arXiv]

We propose a novel framework consisting of two components: Target-Guided and Cycle-Free Data Augmentation and Self-Ensembling algorithm

Self-Training and Adversarial Background Regularization for Unsupervised Domain Adaptive One-Stage Object Detection
Seunghyeon Kim, Jaehoon Choi, Taekyung Kim, Changick Kim
ICCV, 2019   (Oral)
[arXiv]

We introduce a weak self-training (WST) method and adversarial background score regularization (BSR) for domain adaptive one-stage object detection.

Pseudo-Labeling Curriculum for Unsueprvised Domain Adaptation
Jaehoon Choi, Minki Jeong, Taekyung Kim, Changick Kim
BMVC, 2019
[arXiv]

We propose a pseudo-labeling curriculum based on a density-based clustering algorithm.

Service


Conference Reviewer CVPR 2020, WACV 2021, ACCV 2020, AAAI 2021, ICRA 2021

Chosen as one of 66 outstanding reviewers of ACCV 2020
EE838
Teaching Assistant, CMSC733: Computer Processing of Pictorial Information Fall 2022

Teaching Assistant, CMSC250: Discrete Structure Fall 2021

Teaching Assistant, CMSC426: Computer Vision Spring 2021
EE838
Teaching Assistant, EE838–Special Topics in Image Engineering Optimization for Computer Vision Spring 2019

Student Tutor for Foreign Students: Introduction to Programming, CS101 2019

Thanks to Jon Barron! this guy's website is awesome,