StableRep: Synthetic Images from Text-to-Image Models Make Strong Visual Representation Learners.
Yonglong TianLijie FanPhillip IsolaHuiwen ChangDilip KrishnanPublished in: CoRR (2023)
Keyphrases
- visual representation
- image data
- single image
- image features
- input image
- bayesian framework
- object parts
- image analysis
- image content
- random fields
- image classification
- image collections
- image segmentation
- language learning
- image representation
- segmentation method
- image regions
- high resolution
- image retrieval
- expert systems
- pixel values
- multiscale
- object models
- learning environment
- scientific visualization
- graphical user interfaces
- visual representations
- information retrieval
- similarity measure
- web images
- edge detection
- test images
- low level
- probabilistic model