PR-278: RAFT: Recurrent All-Pairs Field Transforms for Optical Flow
The document discusses RAFT (Recurrent All-Pairs Field Transforms for Optical Flow), detailing its network structure, supervised optical flow estimation, and results. It outlines the various datasets used for optical flow, the architecture's features such as energy minimization, and the significance of iterative optimization in estimating optical flow without ground truth. The conclusion highlights the model's impressive performance, while acknowledging a lack of novelty in its optimization approach.
PR-278: RAFT: Recurrent All-Pairs Field Transforms for Optical Flow
1.
RAFT: Recurrent All-PairsField
Transforms for Optical Flow
Hyeongmin Lee
Image and Video Pattern Recognition LAB
Electrical and Electronic Engineering Dept, Yonsei University
6th Semester
PR-278
2.
Content
Datasets forOptical Flow
Supervised Optical Flow Estimation
Network Structure
Training & Inference
Results & Conclusion
Teed, Zachary, and Jia Deng. "RAFT: Recurrent All-Pairs Field Transforms for Optical Flow." arXiv preprint arXiv:2003.12039 (2020). Some Images and Slides From Oral Presentation of this paper: RAFT
Results & Conclusion
Conclusion
• 성능적인 측면에서 봤을 때에는 매우 인상적
• Best Paper치고는, 기술적인 부분과 성능에 Contribution이 치중되어 있다는 사실이 아쉽다.
• Optimization으로 포장하였지만, 사실상 여러 번의 Iteration을 돌린다는 점 외에는 특별히
Optimization이라고 하기는 어렵다.
• 하지만 Optical Flow와 같이, Ground-Truth 없이도 Input을 통해 어느 정도의 평가가 가능한 분야에
서 여러 번의 Iteration을 통한 Inference를 한다는 점은 매우 Reasonable하고 Novel하다고 볼 수
있음