Interpretability of Deep Visual Representations (Paper and Demo)

[siddharth481 (Siddharth)]

Very cool work at CVPR 2017 by students from MIT on dissecting popular CNN models. Would love to know your thoughts.

http://netdissect.csail.mit.edu/

[utk (Utkarsh)]

Looks really cool and much more visually appealing and understandable than the original Matthew Zeiler paper.