Supporting Code for "Self-Supervised Deep Pose Corrections for Robust Visual Odometry"
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
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.
Two bash scripts are provided that will run the training experiments (for monocular pose corrections and stereo pose corrections respectively):
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
This will recompute the pose corrections for a specified KITTI sequence. Currently, it plots the corrected trajectory only.
Reproduction of Paper Results
Our pretrained models are available online. To download them, run the following bash script from the source directory:
paper_plots_and_data, run the various scripts to generate the tables and/or plot the trajectories within our paper. Our pretrained models are also available and can be run on any of the KITTI odometry sequences using