Few-Shot Image Classification Benchmarks are Too Far From Reality: Build Back Better with Semantic Task Sampling.
Etienne BennequinMyriam TamiAntoine ToubhansCéline HudelotPublished in: CoRR (2022)
Keyphrases
- image classification
- visual features
- sports video
- semantic information
- feature extraction
- low level features
- semantic concepts
- random sampling
- benchmark suite
- key frames
- multi label
- semantic network
- image representation
- semantic web
- visual content
- semantic similarity
- semantic content
- semantic knowledge
- parameter space
- domain ontology
- video content
- bag of words
- monte carlo
- natural language
- semantic annotation
- machine learning
- detection algorithm
- domain specific
- domain knowledge
- semantic relationships
- visual concepts
- semantic search
- keywords
- metadata
- knowledge base
- computer vision