Eyes Closed, Safety On: Protecting Multimodal LLMs via Image-to-Text Transformation.
Yunhao GouKai ChenZhili LiuLanqing HongHang XuZhenguo LiDit-Yan YeungJames T. KwokYu ZhangPublished in: CoRR (2024)
Keyphrases
- single image
- image data
- input image
- image content
- multiscale
- image pixels
- image analysis
- multi modal
- image segmentation
- template matching
- high resolution
- image transformations
- feature points
- image collections
- low level
- image matching
- image scrambling
- text graphics
- information retrieval
- lighting conditions
- image retrieval
- hough transform
- image regions
- image features
- edge detection
- spatial information
- access control
- medical images
- text information
- image database