Caption-Generation

File Structure:

• cococaption

• sparsemax

   - core
  
   - data
      - annotations
           - contains json files for training and validation captions
      - test
           - contains pickle files
      - train
           - contains pickle files
      - val
           - contains pickle files
      - imagenet.mat
  
   - data_old (shows what .pkl files were available before training)
  
   - image
           - contains resized images (after preprocessing)
   - jpg
           - contains some results of attention maps
  
   - log (created after training 
           - contains tensorflow logs)
  
   - model
  
   # files
  
   - prepro.py / prepro_small.py
  
   - resize.py
  
   - test.py
  
   - train.py