This project's goal is purely learning and is not applicable to real world scenarios , the data used is extremely limited (208 images total)
In this project , I used a simple CNN , transfer learning on VGG16 and transfer learning on ResNet50. This was a chance for me to use different techniques for the first time like transfer learning , dropout regularization , data augmentation... . Unfortunately , the data was way too limited to draw valuable conclusions from comparaisons