Exploring and Building Visual Question Answering Systems using CLEVR and EasyVQA

Research Report: Exploring and Building Visual Question Answering Systems using CLEVR and EasyVQA

Example 1:

Example 2:

Please follow the below instructions to run our code:

1. Download our best performing model checkpoint from here and place in:

    models/
   

2. Download the CLEVR dataset from here, we've used CLEVR v1.0 Main (Not CoGenT), place the data in:

    CLEVR_v1.0/
    	images/
    	questions/
   

   If you want to instead try it on a simpler, smaller datatset you can try EasyVQA, which has only 13 classes. The code and process remains the same.

3. Run multimodel-clevr-public.py which is in the project root folder: (Make sure all the requirements are installed)

    python multimodel-clevr-public.py