VCR

Visual Commonsense Reasoning

ImagesTextsCustomIntroduced 2019-01-01

Visual Commonsense Reasoning (VCR) is a large-scale dataset for cognition-level visual understanding. Given a challenging question about an image, machines need to present two sub-tasks: answer correctly and provide a rationale justifying its answer. The VCR dataset contains over 212K (training), 26K (validation) and 25K (testing) questions, answers and rationales derived from 110K movie scenes.

Source: Visual Commonsense R-CNN Image Source: From Recognition to Cognition: Visual Commonsense Reasoning

Benchmarks

Explanation Generation/Human Explanation Rating

Related Benchmarks

VCR (Q-A) dev/Visual Question Answering (VQA)/Accuracy VCR (Q-A) dev/Visual Reasoning/Accuracy VCR (Q-A) test/Visual Question Answering (VQA)/Accuracy VCR (Q-A) test/Visual Reasoning/Accuracy VCR (Q-AR) dev/Visual Question Answering (VQA)/Accuracy VCR (Q-AR) dev/Visual Reasoning/Accuracy VCR (Q-AR) test/Visual Question Answering (VQA)/Accuracy VCR (Q-AR) test/Visual Reasoning/Accuracy VCR (QA-R) dev/Visual Question Answering (VQA)/Accuracy VCR (QA-R) dev/Visual Reasoning/Accuracy VCR (QA-R) test/Visual Question Answering (VQA)/Accuracy VCR (QA-R) test/Visual Reasoning/Accuracy