In this work, we introduce the Prototypical Transformer (ProtoFormer), a general and unified framework that approaches various motion tasks from a prototype perspective. ProtoFormer seamlessly integrates prototype learning with Transformer by thoughtfully considering motion dynamics, introducing two innovative designs. First, Cross-Attention Prototyping discovers prototypes based on signature motion patterns, providing transparency in understanding motion scenes. Second, Latent Synchronization guides feature representation learning via prototypes, effectively mitigating the problem of motion uncertainty. Empirical results demonstrate that our approach achieves competitive performance on popular motion tasks such as optical flow and scene depth. Furthermore, it exhibits generality across various downstream tasks, including object tracking and video stabilization.
@article{arxiv.2406.01559,
title = {Prototypical Transformer as Unified Motion Learners},
author = {Cheng Han and Yawen Lu and Guohao Sun and James C. Liang and Zhiwen Cao and Qifan Wang and Qiang Guan and Sohail A. Dianat and Raghuveer M. Rao and Tong Geng and Zhiqiang Tao and Dongfang Liu},
journal= {arXiv preprint arXiv:2406.01559},
year = {2024}
}