AegisPic

Inspiration

I've seen a lot of deep fakes nowadays of popular Celebrities and other people, One day while watching the class Of the course Computer vision And also The paper published by MIT I had an idea And here we are

What it does

Imagine your photo is wearing an invisible forcefield against AI. Here is how it works: The Problem: Bad AI tools can take your face from social media and make fake pictures or clone you. The Magic Trick (Digital Camouflage): We sprinkle tiny, invisible "digital glitter" across your photo. To human eyes, your photo looks completely normal. To an AI robot, that glitter looks like a blinding flashlight. What Happens: When an AI tries to edit your face or make a deepfake, its robot brain gets totally confused and only outputs a blank screen, gray blur, or silly scrambled pixels!

How we built it

"Antigravity"

The Math Engine (PyTorch + PGD) We use an optimization trick called PGD (Projected Gradient Descent). It takes the AI's own neural network and runs it in reverse: "What tiny changes can we spray onto this image so the AI's encoder outputs pure garbage or gray static?"

The Dual Defense PhotoGuard Shield: Targets the AI image generator (Stable Diffusion VAE). If someone tries to inpaint or morph your body/clothes, the AI only outputs gray sludge. Anti-Face-Swap Shield: Scrambles the 512 facial keypoint numbers (ArcFace) so deepfake face-swappers (FaceFusion/RoOP) can't match or paste your face.

The FastAPI Web App A clean web dashboard where you drag-and-drop a photo. You get an interactive split-slider showing the photo looks identical to humans, an amplified 10×10× lens that reveals the invisible digital armor, and a test button showing an AI attack breaking down live.