Use the VQA_LSTM_CNN link to obtain the data. You will need to place the following three files in the data directory --
* 'data/data_prepro.json'
* 'data/data_img.h5'
* 'data/data_prepro.h5'
Write the keras models and place them inside 'models' directory. See the two models provided. simple_mlp is a very basic model for VQA whereas 'DeeperLSTM' is based on VQA_LSTM_CNN
To call your own model pass the argument -model and name of the file. For e.g
python train -model DeeperLSTM