Code for 2019 ICIP paper

This code implemented the method described in the following paper:

Deep Unsupervised Learning for Simultaneous Visual Odometry and Depth Estimation.

2019 IEEE International Conference on Image Processing (ICIP)

To train the network, use:

python3 train.py sfm-learner/KITTI_RAW_DATA/ -b4 -m0.2 -s0.1 --epochs 500 --sequence-length 3 --log-output

#sfm-learner/KITTI_RAW_DATA/# is the path to save the dataset.

To infer the network, use:

python3 run_inference.py --pretrained pretrained_model/Dispnet --dataset-dir test_dir/ --output-dir output_dir/

If you summarize relevant works or refer to the code, please cite:

@inproceedings{lu2019deep,
  title={Deep Unsupervised Learning for Simultaneous Visual Odometry and Depth Estimation},
  author={Lu, Yawen and Lu, Guoyu},
  booktitle={2019 IEEE International Conference on Image Processing (ICIP)},
  pages={2571--2575},
  year={2019},
  organization={IEEE}
}

This implementation is borrowed from SfMLearner paper.