Contrastive Language-Image Pretrained Models are Zero-Shot Human Scanpath Predictors.
Dario ZancaAndrea ZugariniSimon DietzThomas R. AltstidlMark A. Turban NdjeuhaLeo SchwinnBjoern M. EskofierPublished in: CoRR (2023)
Keyphrases
- image analysis
- image data
- multiscale
- input image
- bayesian framework
- random fields
- region of interest
- computational models
- image segmentation
- single image
- image features
- template matching
- programming language
- image classification
- test images
- image retrieval
- static images
- image matching
- hough transform
- image content
- visual effects
- spatial information
- image representation
- natural images
- edge detection
- probabilistic model
- image statistics
- cognitive models
- image pixels
- image regions
- pixel values
- object categories
- image collections
- language learning
- segmentation algorithm
- feature points
- object recognition
- natural language
- similarity measure