ShieldGemma2 for Multimodal Safety

ShieldGemma2 is a safety moderation model for vision language models. It can detect unsafe images. To use it, simply input an image, and provide policies. A policy is a description of what should be detected, and also provide names of policies.

You can test it with an image of your choice and example policies provided below.

Examples
Name of policies separated by semicolon Description of policies separated by semicolon