Stacked Attention Network for VQA

This project was implemented as a part of the course, Image Recognition using Deep learning (Cogs 260) at UC San Diego during Spring 2017. This paper explains the motivation, the algorithm and the results.

Dependencies

1. Tensorflow
2. Skimage
3. Numpy

Dataset download instructions

1. cd data/
2. Download raw data using data_download.sh script
3. Download pickle files of pre-trained model from here

Excecution instructions

python main.py