How to generate synthetic questions and answers, based on the ground-truth annotations of visual concepts?
Closed this issue · 1 comments
LemonQC commented
The author said "We use programs to generate synthetic questions and answers, based on the ground-truth annotations of visual concepts", could you release the code.
Glaciohound commented
Thank you for your interests! VCML/dataset/question_dataset/build_questions/build_questions.py
lists functions for constructing the questions and programs. Hope they can provide information that you are looking for.