Hi all!
I have shared a new notebook with a co-occurence analysis over the labels for all the datasets. I have also made a brief analysis on image similarity using embeddings and PCA.
Have you used any of these ideas? Did you find them useful on your final score?