ViCE: Self-Supervised Visual Concept Embeddings as Contextual and Pixel Appearance Invariant Semantic Representations.
Robin KarlssonTomoki HayashiKeisuke FujiiAlexander CarballoKento OhtaniKazuya TakedaPublished in: CoRR (2021)
Keyphrases
- semantic representations
- visual concepts
- visual similarity
- semantic concepts
- semantic similarity
- contextual information
- visual features
- image annotation
- learning tasks
- image collections
- automatic image annotation
- image content
- visual content
- video content
- semantic gap
- visual data
- object categories
- keywords
- low level features
- visual information
- multi modal
- positive examples
- low dimensional
- image database
- object recognition
- high dimensional data
- wordnet
- web images
- xml documents
- similarity measure