Easy guide to use DeepLabCut on Bowdoin's HPC

This page belongs to the lab of Professor Hadley W. Horch at Bowdoin College. ~~All of her behavioral data, along with ~~DeepLabCut information and user guide are centralized in this GitHub location.

For actual guide on how to use DeepLabCut on Bowdoin's HPC, please refer to Getting-started-DLC.md. Click here to go to the guide.

Want to use existing HPC scripts for your DLC project? Read the section on HPC_Scripts for more information.

References

1. Steps on how to get started using DLC on the HPC (Getting-started-DLC.md)
2. The R scripts and Python files necessary to do so (HPC_scripts)
3. DLC and HPC troubleshooting information (Additional_DLC_Help/*) Some of the information is outdated, but still useful for reference.

Reference to DLC's official github. Reference to the original R script in 2019.

(Deprecated) Link to Tom's forked repository. Some leftover half-finished projects are still going and will be eventually merged back into this repository.

HPC_scripts/*

Scripts that can be used on Bowdoin's SLURM-based HPC Grid for DeepLabCut tracking.

The scripts are designed to work with DLC 2.3.9/2.3.10. As of Jun 2024, the corresponding DeepLabCut version is only installed in the new HPC. So instead of submitting a job on slurm (with ssh slurm.bowdoin.edu), submit those jobs on moosehead (with ssh moosehead.bowdoin.edu). Everything else (sbatch...) should be the same.

To use these scripts:

1. Clone this repository to your HPC folder. Alternatively, download the folder and put it to your HPC/microwave folder.
2. Change necessary parameters in the scripts. Most importantly, the path config.yaml file and paths to your videos.
3. When submitting a job with sbatch, you don't need to add -p gpu ... because the script would request the GPU for you.

R files/*

Post-DLC processing. Visualizing and analyzing movements of the crickets. (This part is Cricket/G. bimaculatus specific).