Shadows Don't Lie and Lines Can't Bend! Generative Models don't know Projective Geometry...for now
This work addresses the problem of detecting AI-generated images for security and verification purposes, though it is incremental as it builds on existing detection methods by focusing on geometric flaws.
The paper demonstrates that generative models produce images with geometric features different from real ones, and classifiers using only geometric properties (perspective fields, lines, and object-shadow relations) detect generated images more reliably than state-of-the-art signal-based detectors.
Generative models can produce impressively realistic images. This paper demonstrates that generated images have geometric features different from those of real images. We build a set of collections of generated images, prequalified to fool simple, signal-based classifiers into believing they are real. We then show that prequalified generated images can be identified reliably by classifiers that only look at geometric properties. We use three such classifiers. All three classifiers are denied access to image pixels, and look only at derived geometric features. The first classifier looks at the perspective field of the image, the second looks at lines detected in the image, and the third looks at relations between detected objects and shadows. Our procedure detects generated images more reliably than SOTA local signal based detectors, for images from a number of distinct generators. Saliency maps suggest that the classifiers can identify geometric problems reliably. We conclude that current generators cannot reliably reproduce geometric properties of real images.