👁️ AI Image Detector Ensemble v2.0

Upload any image and nine independent engines will analyse it to determine if it is AI-generated or real.

# Engine Method
1 FFT Frequency-domain geometric artifact detection
2 ELA Compression-level tampering map
3 Noise Noise pattern forensic analysis (camera fingerprint)
4 ResNet umm-maybe/AI-image-detector
5 SigLIP Ateeqq/ai-vs-human-image-detector
6 SDXL Organika/sdxl-detector
7 ViT-DF prithivMLmods/Deep-Fake-Detector-v2-Model
8 Wvolf Wvolf/ViT_Deepfake_Detection (98.70% acc)
9 SMOGY Smogy/SMOGY-Ai-images-detector (98.18% acc)

Accuracy features: Multi-scale inference (4 views) · Noise forensics · CLAHE enhancement · Adaptive confidence threshold

Android API: POST /analyze with the image as multipart/form-data (field name: file)

🔬 Visual & Forensic Analysis

💡 FFT: Smooth gradient = real. Grid/star patterns = AI.

💡 ELA: Uniform = real. Glowing regions = tampered/AI.

💡 Noise: Bright uniform = real sensor noise. Dark/patterned = synthetic.

🧠 Deep-Learning Scores (% fake — multi-scale averaged)

📦 JSON Result (for API / Android integration)