So i want to make a deep neural network to identify healthy teeth and those with cavities. Making the network is not the problem though, am not quite sure on how i should approach the classification. All i have is a dataset of images with a set of teeth some containing teeth with cavities and some not. So my question is do i use images of a tooth in each image or having the set of teeth will work just fine?
I think you should use all the set of teeth. This is because, I believe, if you train your model with canine teeth only, it will consider all the other set of teeth unhealthy e.g a molar/premolar.
images of a tooth in each image is more like it, I tend to think, you have to think about the impact, cavities are not communicable hence each tooth is treated as is.
@Eric_py might also be right hence it might change, with the question cavities in a set of teeth or cavities in a specific kind of tooth?
For now i want the model to generalize, it needs to classify the given image of a set of teeth as healthy or not healthy despite the number of teeth with cavities in the set. I was thinking to use a set of teeth in each image.
What about using a tooth in each image but have several images of canines, molars, etc in the dataset?
I think that would be better. So that later on you can work on making the model Identify an unhealthy tooth in a picture containing all teeth (like a photo of inside the mouth)