Working on a "special" kind of data (health; medical images etc.) means experiencing the disappointment of realising the method in a paper relies critically on "having a very good image captioner", "asking GPT-4V", etc.

Comments