This app has been built to detect AI-generated images (AIGC) from user-generated images. Currently, it is a limited edition that can detect AIGC images where the main entity in the image is one of the following concepts:
The app has been optimized for AIGC images from Stable Diffusion 1.5, so it is recommended to use images from this model to test AIGC-detector V1.0.
In our internal evaluation, we have observed the following metrics:
We plan to upgrade the app as soon as we are able to figure out ways to meet our cost of R&D and hosting. You can expect the following with tentative timelines:
We have developed another app for detecting data poisoned with state-of-the-art adversarial attack tools such as Nightshade (by UChicago researchers). See the link here & a video demo. You can also review this page for a detailed overview of data poisoning, how it works, mitigation strategies, etc. To know more about the spread of poisoned images online and how artists are using data poisoning to protect their artwork, see here.
Click the button below to start analyzing your images.
Start Detection