Closed sarkarghya closed 2 years ago
This is difficult because the models are trained at a specific input shape. Changing the input shape to the network would change the prediction and it would likely be wrong because the network was not trained that that shape.
There should be some more efficient ways to process a forward pass. Is there a specific issue you are having?
Oh I get it. But then from what I understand some people might not be able to use the model. If there was any website or application which could transform my image to the type that was suitable for the model, it would be really helpful if that were included.
Anyway nice project!
The image should be resized to work accordingly and not give an error.