I created a computer vision model using the image recognition tool. During the training I create a holdout sample to test later for validation against my model, which returns good accuracy and works as expected. What I would like to do however is enter in outside images that are not from my dataset to test on my model. Essentially test how deployable it is against real world samples. For some reason, running the model through the same workflow using my outside data keeps returning the error;
"Unexpected shape (height, width or channel). Make sure the images you input have the same shape as the images you've used to train the mode" .
I am confident all of these factors are the same because I copy pasted the exact same image processing tool I used initially during training as I did to prep my outside images, and am unsure of what else could be causing this issue. Any ideas suggestions would be much appreciated.
The issue was in my choosing to "lock" the resolution, where I thought I was locking in my new specified resolution that I had entered, it was actually keeping the same ratio as my original images causing the error.