Diffusion Lens: Interpreting Text Encoders in Text-to-Image Pipelines.
Michael TokerHadas OrgadMor VenturaDana AradYonatan BelinkovPublished in: CoRR (2024)
Keyphrases
- image data
- multiscale
- information retrieval
- single image
- text information
- web images
- image retrieval
- image features
- image classification
- text retrieval
- textual information
- image analysis
- keywords
- scanned documents
- input image
- text mining
- image content
- textual descriptions
- image collections
- image segmentation
- visual information
- edge detection
- low level
- text documents
- document images
- image representation
- color images
- high resolution
- tensor field
- similarity measure
- text graphics