This project is an image classification project based on a transfer learning approach using with MobileNetV2 architecture.
This project is an improvement on a previous project in which we built and trained a custom deep CNN from the ground up.
Finally, this project demonstrates the power of transfer learning; in fact, we achieved a model accuracy of 98.51% using only 50% of the train dataset, whereas the previous project only achieved 94% accuracy using all available data.
You can find a link provided by microsoft to the dataset used in ths project here .
You can find the model weights and history logs here
 |
 |
 |