To run this project, please get all the following requirements:
- Download the pre-trained word embedding and SNLI dataset or my drive and update data links in the .config files
- 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.
- 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 ...