XAI-SCAN

This Project is working on eXplainable AI (XAI) for Clustering utilizing SCAN and Introspection approaches.

SCAN Clustering:

SCAN was imported from here
The SCAN README file can be found here

XAI Methods:

Introspection Implementation:

GradCam: https://medium.com/@stepanulyanin/implementing-grad-cam-in-pytorch-ea0937c31e82
FeatViz: https://distill.pub/2017/feature-visualization / https://colab.research.google.com/github/tensorflow/lucid/blob/master/notebooks/feature-visualization/regularization.ipynb#scrollTo=gvF6j5h4GkDe

Cluster Word-Cloud:

Generate Image Descriptions with CLIP from OpenAI: https://huggingface.co/docs/transformers/model_doc/clip
Find most frequent Terms and Generate Word-Clouds: https://towardsdatascience.com/generate-meaningful-word-clouds-in-python-5b85f5668eeb (maybe Projection on GradCam)