---
title: AI Image Detector
emoji: 🎥
colorFrom: blue
colorTo: indigo
sdk: streamlit
sdk_version: 1.51.0
app_file: src/app.py
pinned: false
license: apache-2.0
short_description: Detects AI-generated images using ResNet50.
---

# AI-Generated vs. Real Image Detector

### *Can you trust what you see?*

## 👥 Authors

* **Ignacio Alarcon** [LinkedIn](https://www.linkedin.com/in/ignacioalarcon/)
* **Bernardo Gandara** [LinkedIn](https://www.linkedin.com/feed/)

## Demo
![App Demo GIF](/src/demo.gif)

## 🧠 Project Motive & Impact
The rapid advancement of Generative AI has democratized creativity, but it has also introduced significant risks. As synthetic media becomes indistinguishable from reality, the potential for misinformation, identity theft, and digital fraud increases.
We developed this project to address the **"Verification Gap"**—the technological lag between the ability to create fake images and the ability to detect them.

### Why ResNet50?
While human eyes often look for obvious flaws (like extra fingers), AI models like **ResNet50** analyze pixel-level distributions and compression artifacts that are invisible to humans. By using Transfer Learning, we adapted this powerful architecture to distinguish the subtle "digital fingerprint" left by Generative Adversarial Networks (GANs) and Diffusion models.

## 📉 Model Performance
* **Architecture:** ResNet50 (Transfer Learning)
* **Training Data:** Balanced dataset of 100k+ organic and synthetic images.
* **Focus:** Identifying high-frequency artifacts and texture inconsistencies.