paper-with-me

VideoInstruct

Video Instruction Dataset

홈페이지 · 논문 30편

Video Instruction Dataset is used to train Video-ChatGPT. It consists of 100,000 high-quality video instruction pairs. employs a combination of human-assisted and semi-automatic annotation techniques, aiming to produce high-quality video instruction data. These methods create question-answer pairs related to 1. Video summarization 2. Description-based question-answers (exploring spatial, temporal, relationships, and reasoning concepts) 3. Creative/generative question-answers

VideosTexts English

벤치마크

Video-based Generative Performance Benchmarking (Correctness of Information) on VideoInstruct 결과 72개
Video-based Generative Performance Benchmarking on VideoInstruct 결과 69개
Video-based Generative Performance Benchmarking (Consistency) on VideoInstruct 결과 54개
Video-based Generative Performance Benchmarking (Contextual Understanding) on VideoInstruct 결과 54개
Video-based Generative Performance Benchmarking (Detail Orientation)) on VideoInstruct 결과 54개
Video-based Generative Performance Benchmarking (Temporal Understanding) on VideoInstruct 결과 54개
VCGBench-Diverse on VideoInstruct 결과 6개