I tried to use Cats and Dogs notebook for facial recognition using the pictures of John F Kennedy, John Lennon, and John Wayne. The highest accuracy I got was ~40%. Is this the wrong model to use or is it because of my data? I saw a couple of threads about Siamese Network. So maybe, I’m using the wrong one.
Training Set. 10 pictures of each
Validation Set. 3 pictures of each
Test Set. 5 pictures of each
The pictures contains: young and old; with eyeglasses or none; with hats or no hats; long hair or medium hair.
What type of images are applicable for CNN/resnet34?