AttentiveVQA

To run, place datafiles in the following directories

data
└── Questions
    ├── v2_OpenEnded_mscoco_test2015_questions.json
    ├── v2_OpenEnded_mscoco_train2014_questions.json
    └── ...
└── Images
    └── mscoco
        ├── merged2014
        └── test2015
└── Annotations
    ├── v2_mscoco_train2014_annotations.json
    └── v2_mscoco_val2014_annotations.json

Usage

1. python preprocess.py to preprocess comments
2. python generate_features_coco.py to generate image features from the mscoco dataset
3. python train_multimodal.py