IJCAI · 2025
Online 3D Gaussian Splatting Modeling with Novel View Selection
Online 3D Gaussian splatting with novel view selection for building scene models as new observations arrive.
Khang Truong Giang
AI Researcher Ph.D., KAIST
I am an AI researcher focused on reconstructing and understanding the 3D world from images. My work spans 3D reconstruction, neural rendering, and autonomous driving.
I received my Ph.D. in Computer Science from KAIST in 2024, advised by Prof. Sungho Jo and Prof. Soohwan Song. My doctoral research explored multi-view stereo, feature matching, and visual localization. My publications include work at CVPR, ICLR, AAAI, and IJCAI, and in IEEE TIP.
Alongside my academic research, I bring experience in perception and planning for autonomous driving. Previously, I was a 3D Vision Engineer at 42dot, Hyundai Motor Group, where I worked on multi-camera perception. See an autonomous driving demo ↗
Research interests
- 3D reconstruction
- Perception
- Autonomous driving
News
- Our paper on online 3D Gaussian splatting with novel view selection was accepted to IJCAI 2025.
- MVS-GS, our online multi-view stereo mapping work, was published in IEEE Access.
- I joined 42dot (Hyundai Motor Group) as a 3D Vision Engineer.
Earlier updates
- Received the Outstanding Reviewer Award at CVPR 2024.
- DeViLoc was accepted to CVPR 2024 for an oral presentation — top 3.3% (90/2,719 accepted papers).
- Achieved a Silver Medal (top 9%) in the Kaggle Image Matching Challenge.
- Two papers accepted: one at AAAI 2023 and one in Pattern Recognition.
Publications
Full list on Scholar ↗Selected papers, ordered by year.
* Equal contribution · † Corresponding author · Download all citations
CVPR · 2024 Oral presentation · top 3.3% (90/2,719 accepted papers)
Learning to Produce Semi-dense Correspondences for Visual Localization
Semi-dense image correspondences for estimating camera pose and connecting images to a 3D scene.
IEEE Transactions on Image Processing (TIP) · 2024
TopicFM+: Boosting Accuracy and Efficiency of Topic-Assisted Feature Matching
Topic-assisted feature matching designed to improve both matching accuracy and efficiency.
AAAI · 2023
TopicFM: Robust and Interpretable Topic-assisted Feature Matching
Topic-assisted feature matching that connects local correspondences with interpretable scene context.
Pattern Recognition · 2023
Prior Depth-based Multi-view Stereo Network for Online 3D Model Reconstruction
Prior depth information guides multi-view stereo for online 3D model reconstruction.
ICLR · 2022
Curvature-guided Dynamic Scale Networks for Multi-view Stereo
Curvature-guided dynamic scales for estimating depth from multiple views.
RA-L · 2021
Sequential Depth Completion With Confidence Estimation for 3D Model Reconstruction
Sequential depth completion with confidence estimates for 3D model reconstruction.
Service & recognition
CVPR Outstanding Reviewer Award
Recognition for academic peer review at CVPR 2024.
June 2024
Image Matching Challenge · Silver Medal
Top 9% in the Kaggle Image Matching Challenge.
June 2023
Let’s connect
I’m currently on the job market and welcome conversations about opportunities in 3D vision, neural rendering, and autonomous driving. I’m also happy to discuss research collaborations. View my CV (PDF) and get in touch.
khangtg.hust.it@gmail.com ↗