CV

Education

2023/06-2023/11:
- Sensetime Japan
- Duties included: As a researcher intern, we developed VSRD, a novel method for monocular 3D objectdetection with weak 2D supervision, avoiding the need for3D labels. Our approach utilizes multi-view 3D auto-labeling to generate pseudo labels for training. We introduce an Instance-aware volumetric silhouette rendering to create instance masks from a signed distance field (SDF) representation of objects. For optimizing 3D bounding boxes directly, we decompose each object’s SDF into a basic cuboid SDF and a residual distance field (RDF), enabling end-to-end optimization by aligning rendered and actual instance masks. The Paper has been accepted for Conference of Computer Vision and Pattern Recognition.(CVPR2024)
- Supervisor: Hiroki Sakuma
2022/02-2022/04:
- Megvii Shanghai Research Institue.
- Focus on Parking Slot detection and segmentation for autonomous driving.Add a angle constrain to the corners of the parking slot which improve the recall rate of parking slot detection by 10%.Impose a novel training strategy for data augmentation which make full use of the limited training data, which greatly improve the detection at ill-position like image edges and occluded regions. It was finally accepted by Megvii as the IPM pipeline for automatic parking algorithm.
- Supervisor: Zhao Yang

VSRD: Volumetric Silhouette Rendering for Weakly Supervised 3D Object Detection.
Z.Liu*, H.Sakuma*, M.Okutomi
Proceedings of the IEEE/CVF Computer Vision and Pattern Recognition. (CVPR 2024)
* is the equal contribution.
[project page] [paper] [code]
CFDNet: A Generalizable Foggy Stereo Matching Network with Contrastive Feature Distillation.
Z. Liu, Y. Li, M.Okutomi
IEEE International Conference on Robotics and Automation (ICRA2024)
[project page] [paper] [code]
Global Occlusion-Aware Transformer for Robust Stereo Matching.
Z. Liu, Y. Li, M.Okutomi
Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision(WACV 2024)
[project page] [paper] [code]
Digging Into Normal Incorporated Stereo Matching.
Z. Liu, S. Zhang, Z. Wang, M. Okutomi
Proceedings of the 30th ACM International Conference on Multimedia(ACM MM 2022)
[project page] [paper] [code]