Visual Genome

홈페이지 · 논문 1,256편

Visual Genome contains Visual Question Answering data in a multi-choice setting. It consists of 101,174 images from MSCOCO with 1.7 million QA pairs, 17 questions per image on average. Compared to the Visual Question Answering dataset, Visual Genome represents a more balanced distribution over 6 question types: What, Where, When, Who, Why and How. The Visual Genome dataset also presents 108K images with densely annotated objects, attributes and relationships. Source: RaAM: A Relation-aware Attention Model for Visual Question Answering Image Source: Visual Genome: Connecting Language and Vision Using Crowdsourced Dense Image Annotations

ImagesTexts

벤치마크

Unbiased Scene Graph Generation on Visual Genome 결과 93개

Scene Graph Generation on Visual Genome 결과 57개

Object Detection on Visual Genome 결과 20개

Multi-label Image Recognition with Partial Labels on Visual Genome 결과 12개

Layout-to-Image Generation on Visual Genome 128x128 결과 5개

Dense Captioning on Visual Genome 결과 4개

Layout-to-Image Generation on Visual Genome 256x256 결과 4개

Layout-to-Image Generation on Visual Genome 64x64 결과 4개

Visual Relationship Detection on Visual Genome 결과 4개

Phrase Grounding on Visual Genome 결과 3개

Image Generation from Scene Graphs on Visual Genome 64x64 결과 2개

Unsupervised KG-to-Text Generation on VG graph-text 결과 2개

Unsupervised semantic parsing on VG graph-text 결과 1개

Visual Question Answering (VQA) on Visual Genome (pairs) 결과 1개

Visual Question Answering (VQA) on Visual Genome (subjects) 결과 1개