Model for classifying one large object based on small details that are on the object

Deep Learning
1

Hi,

I have a dataset consisting of large images that contain one large single object, this object has small details on it and i want to classify this object into 6 different classes based on the small details that the object has on it. I do not know where exactly the small details are in the large object but they are visible. I’m looking for a model/vision transformer that can help me classify it, I’ve tried using Resnet50 and the results i got were not that good.
Any suggestions which model i should use?

2

I’m afraid you need to be a little more specific. It sounds like a simple mutliclass classification problem to me, but that doesn’t say anything about the reasons why it doesn’t work that well. A resnet50 seems to be a good start…

3

One source of issues could be images significantly larger than the size normally accepted by Resnet, with details disappearing when resizing for Resnet. Maybe transformers might be better suited.