Extract the 4096 dimension activations from the VGGNet built in Keras. The last two layer have been cut off - The last Softmax Layer and Dropout layer.
This has the following dependencies:
- Keras
- Theano
- OpenCV
- Numpy
To run the code :
sudo THEANO_FLAGS=mode=FAST_RUN,device=gpu,floatX=float32 python kerasvgg.py
Link for Pre-Trained Weights : https://www.dropbox.com/s/bvn05hji37lhfoa/vgg16_weights.h5?dl=0