cifake

Deep learning-based image classification system to detect AI-generated synthetic images using CNN and Grad-CAM explainability.

이미지의 AI 생성 여부를 분류하는 서비스

✦ Real-Time AI-Powered Deepfake Detection System ✦ Detect Manipulated Images, Videos & Cloned Voices — Powered by Deep Learning

Arquitetura híbrida clássico-quântica (VQC) para detecção de imagens sintéticas — Brazil Quantum Camp

Sapienza University of Rome - Fundamentals of Data Science 2024/25/1 - Final Project

CIFAKE is an explainable deep learning framework designed to detect AI-generated synthetic images. Built using EfficientNetB0, TensorFlow, Keras, Grad-CAM, and Flask, the system provides accurate image classification along with visual explanations to improve model transparency and user trust.

The Image Authenticity Detector is a machine learning project designed to classify images as either real or AI-generated using a Generative Adversarial Network (GAN)

Cheap, CPU-only AI-image detector (no GPU/PyTorch). A reproducible classical image-forensics pipeline with an optional, still-CPU CLIP (ONNX) neural probe — calibratable, benchmarked, and built to fold in stronger models as they get cheap. CLI, FastAPI, Docker.

A Convolutional Neural Network trained to discriminate between AI-generated and real images.

Detects AI-generated vs real images using a multi-domain deep learning approach — combining RGB, gradient (Sobel), and FFT features with a ResNet18 backbone. Achieves 95.05% accuracy on the CIFAKE dataset. Includes Grad-CAM explainability visualizations.

Robustness evaluation of CNN and Transformer detectors for AI-generated image detection under real-world distribution shifts.

Image classification project using ResNet101 to distinguish real and AI-generated images on the CIFAKE dataset.

💡 Custom PyTorch CNN achieving 96.34% accuracy detecting AI-generated images vs real photography using CIFAKE dataset. Includes live Flask web app for predictions.