You can also find my articles on my Google Scholar profile.

RepPoints V2: Verification Meets Regression for Object Detection
Yihong Chen, Zheng Zhang, Yue Cao, Liwei Wang, Stephen Lin, Han Hu
Neural Information Processing Systems (NeurIPS), 2020
[paper] [code]

Verification and regression are two general methodologies for solving problems while each has its own strengths: verification is easier to infer accurately, and regression is more efficient and applicable to continuous target variables. So how to efficiently combine them to take the merits of the two worlds to improve model performance is an interesting problem. In the field of object detection, methods based on verification (e.g. CornerNet) and regression (e.g. FCOS) all demonstrate promising results but behaves quite differently in detail. Motivated by this interesting phenomena, we make a thorough study about verification and regression, and what we find guides us to build a strong object detector. Our observation could also be extended to instance segmentation and other tasks. We take a step closer to understand neural network's behavior with a new perspective.

Memory Enhanced Global-Local Aggregation for Video Object Detection
Yihong Chen, Yue Cao, Han Hu, Liwei Wang
IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2020
[paper] [code]

Video has now become the most prevalent media format around the world. Thus video object detection, which aims to accurately recognize and locate objects in a piece of video, has great application and research value. Motivated by how human recognize objects in videos, we design an efficient aggregation method which enables our model to utilize far richer information than any previous methods while not introducing heavy computation overheads. We achieve currently the best performance in the challenging ImageNet VID dataset.