Image Classification problem for minute defect detection

I am tasked with the problem of finding defects in a compressor wheel.Here is how a good wheel looks like:

Here is how a defective wheel looks like ( I have drawn a box around the defective area):

I have continuous video feed of the wheels rotating as a data set. I tried training the goodness of a wheel using a fasterrcnn_resnet50_fpn model in pytorch. But the results were inaccurate. This is what I fed in the training data with bounding boxes:

What is the best approach in solving this classification problem? Any hint would be useful.