Transfer learning using Keras

Using the pretrained model of VGG16, we build this model. For a more deatiled understanding check the blog on S/R

Requirements

Python 3.5 Keras 2.0.4 Numpy 1.12.0

Pre-processing the data

Run the data.sh bash file

bash data.sh

This command is used to create the folder necessary for training.

python data.py

This command donwloads the data and performs the pre-processing.

The files are renamed in a chronlogical order. Then they are segmented into data/train, data/test, data/validation Then the labels are saved as pickle.

Training the model

The VGG16 model with pre-trained weights of imagenet is trained on the dataset Check the blog for further understanding.