Evaluating and replicating paper results

[kumam92]

Hi,

thanks for creating such interesting dataset. I wanted to replicate the results in the paper but not sure how to use this repo for that. Can you please guide me for that. I have already downloaded all datasets and class files. By running python run_program.py --qa_dir your_path we can get the output but not sure how training of model and weight file generation is happening. Also do we need to generate scene graph and QA template by ourselves as I thought that generation code is for new question-answer addition.

Thanks!