Suggestion on a usecase
Hello all,
I had a task to do, I didn't understand where to start, Need suggestions from Experts
UseCase: Given a Website, there to register as a user they have to upload few scans, say travel documents.
so my task is to check the the scans are correctly crop and good quality of scan, and also need to match the documents are correct, then only the upload has to successful
Any suggestions will be helpful.
I am asking help for only suggestions to how to start
Disclaimer: I am no expert. Well you wont get a recipe for your project here - its best if you ask specific question in this forum(read the faq).
Before you start: Define the problem and a metric how to measure how good you are. For example object detection - one metric could be MAP. This is basically measuring how much the detected box overlaps with the real thing(called ground truth).
If you do that - things will answer them self. Things like "correctly crop " / "good quality of scan" are too generic to measure. Again - define your problem and a metric on how to measure how you perform on that problem. I know its annoying and costs time - but believe me - in computer vision you better work this way - otherwise you will get lost in the complexity.