VQA-HAT Intelligent Question Answering Dataset

VQA-HAT, also known as VQA Human Attention, can evaluate the information area of an image based on the questions asked. The Cornell University team obtained a human visual attention map by asking volunteers to answer questions about sharpening and blurring areas of the image.

This dataset consists of human visual attention maps on images in the original VQA dataset. It contains more than 60k human visual attention maps.