Image Captioning & Style Transfer

Image captioning with vanilla RNNs and LSTMs on the COCO dataset. Network visualizations and style transfer on SqueezeNet, which is a deep CNN pretrained to perform image classification on ImageNet.

• Vanilla RNN Captioning
• LSTM Captioning
• SqueezeNet Network Visualizations:
  • Saliency Maps:
    
    
  • Fooling Images:
    
    
  • Class Visualization:
    
    
• Style Transfer:
  • Style Transfer:
    
    
  • Feature Inversion:
  

Installation:

sudo pip install virtualenv
virtualenv -p python3 .env
source .env/bin/activate
pip install -r requirements.txt
pip install -r requirements_tf.txt

Download Data:

cd deeplearning/datasets/
./get_data.sh