Researchers from Stanford University have released a dataset of 113,000 images and 22 million related questions to advance computer vision technology. The images are of everyday scenes, such as a table set for breakfast. Accompanying each image are questions that test an AI system’s ability to recognize objects, use spatial reasoning, and make logical inferences. For example, one of the questions for a picture of breakfast food on a table asks “Is the syrup to the left of the napkin?”
Improving Visual Reasoning in AI Systems
