DeepLabCut

Note 220316: See new materials available on the DeepLabCutTM website. 

We use the DeepLabCut (DLC) GUI to label points of interest (PoI) in the grasp environment -- areas of the infant hand as well as reference points such as the ends of the sensorized toy bar.  Persons doing labeling install DLC and its supporting environment (i.e. Anaconda) on their own laptops.  Instructions for installing and invoking DLC follow.  (See "Download and Install Anaconda" below.)

We then use the labeled data to training a deep learning algorithm to recognize points of interest, aiming for a <3 pixel error. To do this, we upload the labeled data file to Google Colab and train there. Instructions for this part of the process are on the Google Colab page.

 See the second file below for a thorough treatment of our approach to labeling points of interest in DLC  locally and training deep learning algorithms using GPUs on Google Colab. 

Download and Install Anaconda

DLC Labeling Protocol Using Locally Installed GUI and Colab for GPUs

(You must be logged in to CV@hugs-lab.org to view this file.)