Images, labels, models and code are provided at ftp://ftp.jax.org/wottoj  for the DeeplabCut model and GentleBoost classifier.  Read_me document describes each file.

Files included

DeepLabCut Model

Zipped file containing the images used to train DeepLabCut and the associated human annotated labels in excel spreadsheet

The yaml output from DLC after training completed

The test yaml output from DLC after training completed

The images with labels from human and model superimposed for train and test images.

GentleBoost Classifier Model

The output of DLC for each of the 40 training videos in H5 files.  Each frame of video is labelled with 159 points representing the x,y coordinates and likelihood values for 4 mice, 12 points per mouse, and 5 points on the arena

The classification of licking and no licking for each frame of video for one arena for each of the 40 videos. Arena indicated by file name.

Matlab structure of the GentleBoost classifier

Matlab code to create the necessary input to classifier.  Code is described using one file of  H5 output of the DLC model.  The output of the classifier is of licking or no licking for each video frame.  Data is typically binned to determine number of seconds of licking in a 5 minute bin.

UPDATE ON FTP DATA LOCATION Aug 2024

Web browsers no longer support connecting directly to FTP servers anymore. Users will need to use different software, such as filezilla, to connect & download. The data is still shared on JAX FTP servers.

If you are still having issues with data access, it is most likely that your institution blocking untrusted ftp sites. JAX’s FTP does not support TLS, which some ITs may consider a security vulnerability because login info is sent unencrypted.

One solution is to access the data from a place that hasn’t restricted it (home or public internet).  The other is to ask your IT folks to enable access. The third is for us to move the data to another location (can would require an edit to the published paper).