image-caption-generator stores the encoder, decoder and attention layers of the model based on the architecture discussed in Xu et al.'s Show, Attend and Tell: Neural Image Caption Generation with Visual Attention. train.py trains the model based on global hyperparameters and then produces a sample caption.
The analysis folder stores some files created by us and used to deploy Sheng et al.'s sentiment and regard classifiers on our generated captions.
Project by Jiin Kim, Arjun Kallapur, Aman Oberoi.