Self-Supervision on Images and Text Reduces Reliance on Visual Shortcut Features.
Anil PalepuAndrew L. BeamPublished in: CoRR (2022)
Keyphrases
- image features
- visual appearance
- original images
- web images
- extracting features
- salient features
- automatically extracted
- input image
- test images
- visual objects
- low level
- visually similar
- image data
- image database
- object recognition
- image analysis
- visual attributes
- visual information
- visual patterns
- semantic content
- image registration
- textual information
- image matching
- keypoints
- visual features
- content features
- visual similarity
- image set
- image annotation
- image retrieval
- feature set
- edge detection
- image similarity
- visual data
- spatial layout
- extracted features
- sample images
- visual descriptors
- image collections
- multiple modalities
- spatial information
- image regions
- semantic information
- co occurrence
- keywords
- low level visual features
- visual scene
- image search
- text information
- visual content
- feature descriptors
- feature extraction
- spatial relationships
- image classification
- complex background
- gabor filters
- image content
- feature vectors
- observed scene