Hi just wonder are there formula to determined how many convolution layer should there be for CNN im not sure if i put too much that will be over fit?
I’m not sure that anyone can really answer that for you, there are all types of architectures with all types of use cases, an no formula that will give you an exact answer as to how much you should use. With that said, you should probably look at the commonly used networks which are available for fine-tuning in pytorch to get a feel for what is standard. You can find them at
https://pytorch.org/docs/stable/torchvision/models.html and you can import them as shown in that page in the docs. You can then just print out the model and it will print out it’s architecture.
thank you so much will check it out