FakeInversion: Learning to Detect Images from Unseen Text-to-Image Models by Inverting Stable Diffusion.
George CazenavetteAvneesh SudThomas LeungBen UsmanPublished in: CoRR (2024)
Keyphrases
- input image
- image data
- image features
- image collections
- test images
- image classification
- image retrieval
- image analysis
- original images
- auto annotation
- image set
- web images
- image pixels
- pixel values
- region of interest
- complex background
- random fields
- edge detection
- image regions
- image statistics
- lighting conditions
- object detectors
- unlabeled images
- sample images
- object recognition
- feature points
- segmentation method
- image content
- linear predictors
- image database
- image matching
- single image
- image structure
- labeled images
- bounding box
- multiscale
- normalized gradient
- synthesized images
- image quality
- segmentation algorithm
- image segmentation
- contrast enhancement
- secret image
- natural images
- set of training images
- image search
- scale space
- textual information
- image blocks
- textual descriptions
- high resolution
- tensor field
- training set
- visual features
- image representation