Diffusion Lens: Interpreting Text Encoders in Text-to-Image Pipelines.
Michael TokerHadas OrgadMor VenturaDana AradYonatan BelinkovPublished in: ACL (1) (2024)
Keyphrases
- image data
- single image
- text information
- information retrieval
- multiscale
- keywords
- image classification
- image features
- image representation
- text retrieval
- image segmentation
- textual and visual information
- wide angle
- scanned documents
- textual information
- image collections
- input image
- web images
- image analysis
- image retrieval
- hough transform
- segmentation algorithm
- edge detection
- text mining
- handwritten words
- text graphics