Open-DDVM: A Reproduction and Extension of Diffusion Model for Optical Flow Estimation
This work provides an open-source alternative to a closed-source model for optical flow estimation, making it more accessible to researchers.
The authors reproduced and extended Google's DDVM diffusion model for optical flow estimation, achieving comparable performance to the closed-source original by training on 40k public data with 4 GPUs.
Recently, Google proposes DDVM which for the first time demonstrates that a general diffusion model for image-to-image translation task works impressively well on optical flow estimation task without any specific designs like RAFT. However, DDVM is still a closed-source model with the expensive and private Palette-style pretraining. In this technical report, we present the first open-source DDVM by reproducing it. We study several design choices and find those important ones. By training on 40k public data with 4 GPUs, our reproduction achieves comparable performance to the closed-source DDVM. The code and model have been released in https://github.com/DQiaole/FlowDiffusion_pytorch.