based on the set of 789 images of pills, a machine learning algorithm was trained. The animation runs for 33 sec and displays a latent space walk through the vector field.
All initial pictures were taken in course of the project "C3PI: Computational Photography Project for Pill Identification" (Bodenreider, Yaniv). The U.S. NLM used the images for the Pillbox project, but retired it in Jan 2021.
The entire database for the U.S. consisted of 4447 different pills up to this point.