Hello!
I’m creating a dataset for screws, but there are pictures like this where they overlap with each other. Should I label them as one if my goal is to check if the images has a screw or not.
what is this for ? what is the goal of your machine learning application later ?
you will need train images, that are similar to your actual inference situation (and glossy stock photos like above are probably a bad idea)
if there can be several screws in your images and you want to classify those as “has screws”, it’s ok like above, if you want “contains a single screw”, then ofc. not.
1 Like
Basically my program has to detect if an image contains the bolt or not. There can be several of them in the image, but I don’t have to count them.
Another question, can these “glossy” photos affect the accuracy if the image I’m trying to detect on is pretty dark?