This repo contains code for analysing behavioral and neural data. The two primary components are the behavenet package for explicit and latent behavioral analysis, and the notebooks folder, which contains jupyter notebooks to visualize the results.

Behavenet requires data be formatted in hdf5 files to fit its models. For a detailed overview of how these hdf5 files are expected to be organized, you can refer to the behavenet documentation.

For our purposes, we'll need the following data/files:

• a .mj2 video of mouse behavior (for these experiments we expect a video of a mouse's eye)
• a file containing the corresponding neural data (neurons x time)
• two files that contain indexes to line up video and neural data (2 x num_trials, where batch_idxs[0][i] indicates start index of a trial and batch_idxs[1][i] indicates the end index)
• behavioral data - a .npy file extracted from facemap (for the time being - this may switch to deeplabcut extracted labels soon)
• a file that contains information about individual cell types, and good/bad cells

The most relevant piece of code is scripts/data_to_hdf5.py script, which formats the data into the required format. This code expects data to be organized in the following directory structure: - root_folder - lab - date - cell_info_for_mouse