Supporting Code for "Self-Supervised Deep Pose Corrections for Robust Visual Odometry"

Dependencies:

• numpy
• scipy
• pytorch Tested with 1.4.0 (Python version 3.7.4)
• liegroups
• pyslam
• tensorboardX

Datasets

We trained and tested on the KITTI dataset. Download the raw dataset here. We provide a dataloader, but we first require that the data be preprocessed. To do so, run create_kitti_data.py within ss-dpc-net/data (be sure to specify the source and target directory). We preprocessed the data by resizing the images and removing 'static' frames.

Training

Two bash scripts are provided that will run the training experiments (for monocular pose corrections and stereo pose corrections respectively):

run_mono_exps.sh

run_stereo_exps.sh

Prior to training, the data directory should be modified accordingly to point to the processed KITTI data. During training, to visualize the training procedure, open a tensorboard from the main directory:

tensorboard --logdir runs

Pretrained Models

Our pretrained models are available online. To download them, run the following bash script from the source directory:

bash download_data.sh

Inference

run:

run_inference.py

This will recompute the pose corrections for a specified KITTI sequence. Currently, it plots the corrected trajectory only.

Reproduction of Paper Results

Within paper_plots_and_data, run the four scripts to generate the tables and/or plot the trajectories within our paper.

Citation

If you use this in your work, please cite:

@inproceedings{2020_Wagstaff_Self-Supervised,
  author = {Brandon Wagstaff and Valentin Peretroukhin and Jonathan Kelly},
  booktitle = {Proceedings of the {IEEE} International Conference on Robotics and Automation {(ICRA'20})},
  date = {2020-05-31/2020-06-04},
  month = {May 31--Jun. 4},
  title = {Self-Supervised Deep Pose Corrections for Robust Visual Odometry},
  url = {https://arxiv.org/abs/2002.12339},
  year = {2020}
}