Understanding Misclassifications
Kyle and Ishan discuss the importance of differentiating misclassifications based on the Relevance score and the Presence score, and how it helps in interpreting the output of the classifier. Ishan also mentions a figure in his paper that demonstrates the variation in the relevant score for visually present concepts.In this clip
From this podcast
Data Skeptic
Machine Learning on Images with Noisy Human-centric Labels
Related Questions