We use TensorFlow + Keras to build models

We have 4 models Convolutional neural network (CNN) :

1. CNN (fashion_mnist dataset)
2. CNN (v01) (cats_and_dogs_small dataset)
3. CNN With data augmentation (v02) (cats_and_dogs_small dataset)
4. CNN With VGG-16 With data augmentation (v03) (cats_and_dogs_small dataset)
5. CNN With VGG-16 With data augmentation With Unfreeze the base model (v04) (cats_and_dogs_small dataset)

we find them in CNN.ipynb , Project.ipynb and VGG-16.ipynb

cats_and_dogs_small.zip - Google Drive

This model is cnn_fashion_mnist.h5

we find code in CNN.ipynb => Code view here 👨‍💻

This model is project_V01.h5

we find code in Project.ipynb => Code view here 👨‍💻

This model is project_V02.h5

we find code in Project.ipynb => Code view here 👨‍💻

This model is project_V03.h5

we find code in Project.ipynb => Code view here 👨‍💻

We not change weights in convolutional base VGG-16 ( freeze the base model ) and train only fully connected neural network

VGG-16 Architecture :

This model is project_V04.h5 (download model weights from here )

we find code in VGG-16.ipynb => Code view here 👨‍💻

We change weights in convolutional base VGG-16 ( Unfreeze the base model ) and train fully connected neural network

VGG-16 Architecture :