According to my previous post I had the issues with Google Vision OCR Recognition. So I've trained the AutoML Vision model from scratch using the topic here. I created a two kings of datasets Multi-Label Classification and Single-Label Classification. For both datasets I uploaded the croped images with numbers 0-9 and sybmols B-F. For this images I've attached it's labels (0-9 and B-F). After training ang evaluating I've tried to test it. When I upload and image with single object (number) it works perfect but when I use the image with multiple objects (numbers) then it recognizes it with no objects or recognizes incorrectly. The multiple objects recognition doesn't works in both kinds of datasets. What do I wrong or Google AutoML doesn't support that?
1.4m articles
1.4m replys
5 comments
57.0k users
