Skip to content
code for Stacked attention networks for image question answering
Branch: master
Clone or download
Latest commit 14a8841 Jan 7, 2017
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
data_vqa
expt
src
.gitignore
README.md minor Apr 10, 2016

README.md

imageqa-san

Source code for Stacked attention networks for image question answering.

Joint collaboration between CMU and MSR.

Dependencies

The code is in python and uses Theano package.

  • Python 2.7
  • Theano
  • Numpy
  • h5py

Usage

Download the data from here and extract them at data_vqa folder.

cd src; python san_att_conv_twolayer.py

to start training.

Reference

If you use this code as part of your research, please cite our paper

''Stacked Attention Netowrks for Image Question Answering'', Zichao Yang, Xiaodong He, Jianfeng Gao, Li Deng and Alex Smola. To appear in CVPR 2016.

@article{YangHGDS15,
author    = {Zichao Yang and
Xiaodong He and
Jianfeng Gao and
Li Deng and
Alexander J. Smola},
title     = {Stacked Attention Networks for Image Question Answering},
journal   = {CoRR},
volume    = {abs/1511.02274},
year      = {2015},
url       = {http://arxiv.org/abs/1511.02274},
}
You can’t perform that action at this time.