beexplainable 1.2

Design for HIL experiment to test uncertainty and didactics.

beexplainable 1.1

Release 1.1

• CoProNN: Concept based Prototypical Nearest Neighbors
• Extensive comparisons against TCAV and IBD

beexplainable 1.0

Release 1.0

• Conducted human-in-the-loop experiment via jsPsych to investigate whether explanations help humans to learn relevant rules for classifying wild bees.
• XAI Method based on nearest neighbors w.r.t. to concept images generated with Stable Diffusion.

beexplainable 0.1

Release 0.1

Date: 6.07.2022

😎 New Features

beexplainable package containing following libraries:

• preprocessing: tools for preprocessing images and generating additional datasets based on given or computed annotations
• modelling: ResNet50 constructor
• utils: tools for computing annotations and creating metafiles