The dataset is uploaded to AWS S3, feel free to download and train your models!

  • Train questions S3 URI (s3://mqa-capstone-datasets/ourQA/final/Train.json)

455,622 questions

  • Val questions S3 URI (s3://mqa-capstone-datasets/ourQA/final/Val.json)

155,221 questions

  • Test questions S3 URI (s3://mqa-capstone-datasets/ourQA/final/Test.json)

155,626 questions

QA Schema Example

"621843": 
{"image_id": "VG_100K/2.jpg", 
 "question": "How many cars are parked?", 
 "question_type": "How", 
 "answer": 2}