CV

Education

Libo Long, Xiao Hu, Jochen Lang. “Robustness Analysis: Are Optical Flow Methods Safe to Use? “. Accepted by Conference on Computer Vision and Pattern Recognition (CVPR) 2025.
Xiao Hu, Libo Long, Jochen Lang. “Motion Decoupled 3D Gaussian Splatting for Dynamic Object Representation”. Proceedings of the 39th AAAI Conference on Artificial Intelligence 2025.
Libo Long, Xiao Hu, Jochen Lang. “Learning to Handle Large Obstructions in Video Frame Interpolation”. Proceedings of the 32nd ACM International Conference on Multimedia (ACM MM) 2024.
Xiao Hu, Basavaraj Hampiholi, Heiko Neumann, Jochen Lang. “Temporal Context Enhanced Referring Video Object Segmentation”. Winter Conference on Applications of Computer Vision (WACV) 2024.
Xiao Hu,Jochen Lang. “DOE-SLAM: Dynamic Object Enhanced Visual SLAM”. Sensors 2021, 21(9), 3091.

Research Assistant Spet 2020-Present
- University of Ottawa
- Computer Vision, 4D Reconstruction, SLAM under dynamic environment, RVOS.
Computer Vision Research Intern Spet 2021-Apr 2022
- National Research Council Canada (NRCC)
- Real-world computer vision tasks, including semantic segmentation, pedestrian detection and tracking, and trajectory prediction.
Distributed Systems Intern Spet 2018-Dec 2018
- Wind River Systems, Inc.
- Cloud computing system contributions (based on OpenStack) and automated cloud system installation.
Data Analysis Intern May 2018-Sept 2018
- Environment and Climate Change Canada (ECCC)
- Data analysis and river water levels prediction in the Ottawa region.

M5D-GS 2024
- Developed a 3D Gaussian splatting-based method for 4D dynamic object representation.
- Decoupled object motion from deformation to improve processing of large-motion objects.
- Proposed a novel dataset for large-motion objects.
TCE-RVOS 2023
- Developed a multi-modal approach for referring video object segmentation
- Enhanced the ability to understand temporal context.
Vulnerable Individual-Aided Smart Crossroad System 2022
- Implemented pedestrian detection and classification from surveillance images.
- Estimated and predicted trajectories for vulnerable individuals (elderly, pregnant, and disabled people).
- Incorporated with the traffic light system to control the green light duration.
UAV-Based Railway Inspection 2021
- Segmented railway tracks and water surfaces from UAV images to prevent water overflow onto the tracks.
- Calibrated GPS information to align images from different conditions and seasons, for risk level analysis.
DOE-SLAM 2020
- Estimated the camera pose and object motion simultaneously.
- Recovered the camera pose from object motion during obstruction.
Dynamic Object SLAM 2019
- Based on ORB-SLAM and Mask R-CNN to reconstruct different object models separately, and improve the accuracy in dynamic environments.
Sequence Based SLAM 2019
- Selected the sequence based method with sequence descriptor to reduce the time usage for loop detection.

National Research Council Canada
- Outstanding Intership Award Apr 2022
University of Ottawa
- Hasan Ural Memorial Scholarship Oct 2021
- International Experience Scholarship May 2021
- International Admission Scholarship Aug 2020
Huazhong University of Science and Technology
- Excellent Graduate Award Jun 2017
- Excellent Student Cadre Oct 2016
- Individual Scholarship(Social Welfare) Mar 2016