To run this project, please get all the following requirements:

1. Download the pre-trained word embedding and SNLI dataset or my drive and update data links in the .config files
2. Download Flickr30k data set and features for each image (7x7x512). Or you can get it from my drive: Flickr30k,file names, image features then update the directories in the image_utils.py.
3. Run python main.py --config_file=bimpm_baseline.config

view results: viewer.html is for viewing the prediction results. Since browsers do not allow html and js to access local file, you need to install a web server. I use a simple extension Web server for Chrome. Open this extension and set the folder to be where .result file locate. Update the viewer.html file to view the desired result file.

Prerequisite: tensorflow 1.0.1, numpy, python 2 ...

BiMPM: https://github.com/zhiguowang/BiMPM