@inproceedings{4ab7b3551f8f4337b7a62100692ef0a2,
title = "SEA-RAFT: Simple, Efficient, Accurate RAFT for Optical Flow",
abstract = "We introduce SEA-RAFT, a more simple, efficient, and accurate RAFT for optical flow. Compared with RAFT, SEA-RAFT is trained with a new loss (mixture of Laplace). It directly regresses an initial flow for faster convergence in iterative refinements and introduces rigid-motion pre-training to improve generalization. SEA-RAFT achieves state-of-the-art accuracy on the Spring benchmark with a 3.69 endpoint-error (EPE) and a 0.36 1-pixel outlier rate (1px), representing 22.9\% and 17.8\% error reduction from best published results. In addition, SEA-RAFT obtains the best cross-dataset generalization on KITTI and Spring. With its high efficiency, SEA-RAFT operates at least 2.3× faster than existing methods while maintaining competitive performance. The code is publicly available at https://github.com/princeton-vl/SEA-RAFT.",
author = "Yihan Wang and Lahav Lipson and Jia Deng",
note = "Publisher Copyright: {\textcopyright} The Author(s), under exclusive license to Springer Nature Switzerland AG 2025.; 18th European Conference on Computer Vision, ECCV 2024 ; Conference date: 29-09-2024 Through 04-10-2024",
year = "2025",
doi = "10.1007/978-3-031-72667-5\_3",
language = "English (US)",
isbn = "9783031726668",
series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
publisher = "Springer Science and Business Media Deutschland GmbH",
pages = "36--54",
editor = "Ale{\v s} Leonardis and Elisa Ricci and Stefan Roth and Olga Russakovsky and Torsten Sattler and G{\"u}l Varol",
booktitle = "Computer Vision – ECCV 2024 - 18th European Conference, Proceedings",
address = "Germany",
}