My AI model finally hit 87% accuracy after 3 months, but the weird part was what got me there
I've been messing with a small image classifier for my brother's coin collection since January, and for weeks it was stuck at 61% no matter what I tried. Last Tuesday I was at my kitchen table in Boise at like 11pm and realized I was feeding it way too many blurry pics from my phone. I pulled in 400 clean scans from a flatbed scanner instead, and by Friday it jumped to 87%. What got me was that the fix had almost nothing to do with the model itself, just the data going in. Has anyone else had a jump like that just from swapping your input source?