Automatic Generation of Contrast Sets from Scene Graphs: Probing the Compositional Consistency of GQA.
Yonatan BittonGabriel StanovskyRoy SchwartzMichael ElhadadPublished in: NAACL-HLT (2021)
Keyphrases
- video sequences
- automatically generate
- d scene
- three dimensional
- scene analysis
- scene classification
- graph matching
- image sequences
- scene understanding
- input image
- graph representation
- graph theory
- multiple images
- foreground background
- indoor and outdoor
- graph mining
- dynamic scenes
- single image
- graph model
- complex scenes
- social networks
- camera motion
- image regions
- outdoor scenes
- object recognition
- scene images