This is a Pytorch implmentation for the paper Deep Cross-Modal Projection Learning for Image-Text Matching.
The official implementation in TensorFlow can be found here.

• Python 3.5
• Pytorch 1.0.0 & torchvision 0.2.1
• numpy
• scipy 1.2.1

• Download the pre-computed/pre-extracted data from GoogleDrive and move them to data/processed folder. Or you can use the file dataset/preprocess.py to prepare your own data.
• [Optional] Download the pre-trained model weights from GoogleDrive and move them to pretrained_models folder.

You should firstly change the param model_path to your current directory.

sh scripts/run.sh

You can directly run the code instead of performing training and testing seperately.
Or training:

sh scripts/train.sh  

Or testing:

sh scripts/test.sh