So, for my final project I did a silly model to do something useful: to identify new tires and worn out tires so that one can use this model to check if the tires on his car are worn out and need replacement.
I tried to do a CNN model with not much success: course-project-worn-out-tires
And then I tried to use a ResNet32 model with a little better success: course-project-worn-out-tires-resnet
Still, to get a good and well-trained model I would need to collect and clean up a better dataset with pictures of tires and probably do some improvements in the model too.
In the end I wrote a blog post here: https://medium.com/@ntufar/neural-network-to-identify-if-tires-are-worn-out-and-need-replacement-27b22c612180
Hope it will be useful to someone or inspire someone to do great and better things.