Can anyone help me with this. The model has about a 80 percent accuracy but when I try to send in images for it to predict it would always say that the image is a human even if the image is of a robot, but I would be able to get correct image classifications when I comment out the line of code that would normalize the images by dividing the pixel values by 255.
attached is my notebook file along side the test images and the model weights. I can’t upload my dataset so load the weights found in the test images zip file to test it out. I would like anyone to test it out and if they could explain to me what is going on
human-or-robot-classifier.ipynb (1.9 MB)
test images.zip (26.4 MB)