Bayrambai / raft-msf

RAFT-MSF

This repository is the official PyTorch implementation of the paper:

   RAFT-MSF: Self-Supervised Monocular Scene Flow Using Recurrent Optimizer
    Bayram Bayramli Junhwa Hur and Lu Hongtao

• Contact: bayram.xiaolong[at]gmail.com

Installation and Dataset

For installation and configuring dataset please refer to Self-Supervised Monocular Scene Flow Estimation

Training and Inference

For training, you can simply run the following script file:

./train_monosf_selfsup_kitti_raw.sh

For evaluation the pretrained model (RAFT-MSF), you can simply run the following script file:

./eval_monosf_selfsup_kitti_train.sh

Pretrained Models

The ckpt folder contains the checkpoints of the pretrained model.

Acknowledgement

Please cite our paper if you use our source code.

@article{Bayramli2022RAFTMSFSM,
  title={RAFT-MSF: Self-Supervised Monocular Scene Flow using Recurrent Optimizer},
  author={Bayram Bayramli and Junhwa Hur and Hongtao Lu},
  journal={International Journal of Computer Vision},
  year={2023},
  url={https://doi.org/10.1007/s11263-023-01828-4}
}

The overall code framework adapted from Self-MONO-SF and RAFT. We thank the authors for their contribution.