On the Proactive Generation of Unsafe Images From Text-To-Image Models Using Benign Prompts.
Yixin WuNing YuMichael BackesYun ShenYang ZhangPublished in: CoRR (2023)
Keyphrases
- input image
- image data
- image features
- image collections
- test images
- web images
- original images
- image retrieval
- image regions
- image analysis
- image classification
- image pixels
- image set
- pixel values
- region of interest
- image statistics
- high contrast
- intensity values
- image processing algorithms
- imaging process
- random fields
- image matching
- segmentation algorithm
- sample images
- edge detection
- image database
- image structure
- lighting conditions
- prior models
- segmentation method
- imaging model
- visual effects
- synthesized images
- image content
- gray value
- grey level
- illumination conditions
- text information
- feature points
- contrast enhancement
- image search
- object recognition
- set of training images
- image blocks
- image segmentation
- spatial information
- keypoints
- image annotation
- textual descriptions
- street view
- visual features
- image processing
- scanned images
- text queries
- multiscale
- image registration
- scale space
- natural images
- textual information
- image representation
- digital imaging
- gray level
- secret image
- visual information
- historical manuscripts
- imaging artifacts