Unsafe Diffusion: On the Generation of Unsafe Images and Hateful Memes From Text-To-Image Models.
Yiting QuXinyue ShenXinlei HeMichael BackesSavvas ZannettouYang ZhangPublished in: CCS (2023)
Keyphrases
- input image
- image data
- image features
- image collections
- test images
- web images
- image retrieval
- original images
- image analysis
- image pixels
- image statistics
- region of interest
- image classification
- image regions
- random fields
- pixel values
- image matching
- edge detection
- image set
- image content
- segmentation method
- visual effects
- digital imaging
- synthesized images
- lighting conditions
- street view
- image database
- text queries
- spatial information
- sample images
- image blocks
- imaging process
- text information
- feature points
- segmentation algorithm
- set of training images
- bounding box
- contrast enhancement
- textual information
- text regions
- textual descriptions
- image structure
- secret image
- normalized gradient
- three dimensional objects
- illumination conditions
- natural images
- gray value
- trademark images
- gaussian noise
- anisotropic diffusion
- object recognition
- scanned documents
- medical images
- image sequences
- multiscale
- scanned images
- image registration
- active contours
- keypoints
- mr images
- tensor field
- image search
- image gradient
- text lines