Code and data related to the Neural-Swarm and Neural-Swarm2 (ICRA 2020, T-RO 2021) papers
For training, please run learning/training.py, which will use data in data/training.
Note that learning/training.py also includes data preprocessing, model saving, analysis, and visualization (please see comments therein).
This relies on some data files from Box, to be copied to the data/ folder.
This process can be automated using rclone:
- Setup rclone with box, following https://rclone.org/box/ (note "user" type worked just fine for me)
- To pull the latest data use:
rclone copy box:neural-swarm data/ --dry-run
rclone copy box:neural-swarm data/ -P
- To push the latest data use:
rclone copy data/ box:neural-swarm --dry-run
rclone copy data/ box:neural-swarm -P