A guy at a Cleveland hackathon asked why my model needed 40 labels when 6 would do
I built a image sorter for a local food bank last month. I labeled about 40 different things, canned goods, boxes, produce, all of it. This guy Marcus watched me for 10 minutes and asked one question. Why not just sort into 6 big groups first and let it flag the weird stuff? I cut my labels down and it got way more accurate. Anyone else find that fewer categories made their setup work better?