Sithu Aung.

PhD Student

Sithu Aung

3D vision, geometry & scene understanding

I study 3D vision at Czech Technical University in Prague.

Explore my workGet in touch
Sithu Aung

News & updates

Publication

Depth2Pose at NeurIPS 2026

Our pose-based benchmark for monocular depth estimation appears in the NeurIPS 2026 ED Track.

Publication

PLOT at ECCV 2026

Pseudo-labeling via object tracking for monocular 3D object detection.

Research interests

3D Vision

Recovering geometry, camera pose, and structure from images.

Scene Reconstruction & Understanding

Reconstructing scenes and understanding objects, people, and their spatial relationships.

Digital Twins

Exploring visual representations that connect physical environments with their digital counterparts.

Publications

Research in geometry, depth, and multi-view understanding.

5 PAPERS

* denotes equal contribution.

NeurIPS 2026 · ED Track
Depth2Pose benchmark overview for depth-aware relative pose estimation
3D VisionDepth Estimation

Depth2Pose: A Pose-Based Benchmark for Monocular Depth Estimation without Ground-Truth Depth

Viktor Kocur*, Sithu Aung*, Gabrielle Flood*, Yaqing Ding, Lukas Bujnak, Torsten Sattler, Zuzana Kukelova

A task-driven evaluation framework for monodepth based on depth-aware relative pose estimation, and a new dataset connected to this.

PaperCodeProject
ECCV 2026
PLOT monocular 3D object detection demonstration
Scene UnderstandingObject Detection

PLOT: Pseudo-Labeling via Object Tracking for Monocular 3D Object Detection

Seokyeong Lee*, Sithu Aung*, Junyong Choi, Seungryong Kim, Ig-Jae Kim, Junghyun Cho

A practical training-free pseudo-labeling framework for 3D object detection.

PaperProject
AAAI 2025
Multi-view pedestrian occupancy prediction in MVPOcc
Scene UnderstandingOccupancy Prediction

MVPOcc: Multi-View Pedestrian Occupancy Prediction with a Novel Synthetic Dataset

Sithu Aung, Min-Cheol Sagong, Junghyun Cho

A new occupancy prediction dataset for dense pedestrians in multi-view environments.

PaperProject
WACV 2024
MVFP multi-view pedestrian detection demonstration
Scene UnderstandingMulti-view Detection

MVFP: Enhancing Multi-view Pedestrian Detection Through Generalized 3D Feature Pulling

Sithu Aung, Haesol Park, Hyungjoo Jung, Junghyun Cho

An effective 3D feature pulling method for generalizable multi-view pedestrian detection.

PaperVideo
View Google Scholar

Academic background

Service & reviewing

  • ICCV 2025Reviewer
  • BMVC 2025Reviewer
  • CVPR 2026Reviewer
  • ECCV 2026Reviewer
  • NeurIPS 2026Reviewer

Contact

For questions about my work or research conversations, feel free to get in touch.

Google ScholarGitHub