How kirteshchavhan/forgery works
NeuralForensics AI is a web platform for image-forgery detection, localization, and forensic reporting. The principal workflow is user upload through the dashboard to the API, which evaluates the image with a dual-stream CNN and three image-analysis techniques, fuses their findings, and can produce a PDF report. Separate capabilities include account management, dataset benchmarking, and additional forensic analyses documented in the README. The sampled API imports support its core analysis and report links; the additional techniques are preserved as capabilities, but their API wiring was not shown.
Main entry points: app.py Analysis API · dataset_manager.py Dataset benchmark · app.js Forensics dashboard · auth.py Account authentication
Components 18
Web interface
- Forensics dashboard
frontend/js/app.js
API and accounts
- Analysis API
backend/app.py - Account authentication
backend/auth.py - User records
backend/auth_data/users.json
Forensic analysis
- Dual-stream CNN
backend/models/cnn_detector.py - Grad-CAM localization
backend/models/gradcam.py - Chrominance DCT
backend/forensics/chrominance_dct.py - Micro-texture analysis
backend/forensics/micro_texture.py - Frequency analysis
backend/forensics/frequency_analysis.py - Multimodal fusion
backend/forensics/ensemble.py - Copy-move matching
backend/forensics/copy_move.py - Error-level analysis
backend/forensics/ela.py - Sensor-noise analysis
backend/forensics/noise_analysis.py - Metadata audit
backend/forensics/metadata_analysis.py
Benchmark and reporting
- Dataset benchmark
backend/dataset_manager.py - PDF report generation
backend/forensics/report_generator.py - CNN training
backend/models/trainer.py
Other components
- Analyst
Connections 17
- Analyst → Forensics dashboard: submits image
README.md - Forensics dashboard → Analysis API: uses API
README.md - Analysis API → Account authentication: authenticates users
backend/app.py - Account authentication → User records: reads and writes
backend/auth.py - Analysis API → Dual-stream CNN: runs detection
backend/app.py - Analysis API → Grad-CAM localization: generates heatmap
backend/app.py - Analysis API → Chrominance DCT: analyzes image
backend/app.py - Analysis API → Micro-texture analysis: analyzes image
backend/app.py - Analysis API → Frequency analysis: analyzes image
backend/app.py - Analysis API → Multimodal fusion: fuses findings
backend/app.py - Analysis API → PDF report generation: generates report
backend/app.py - Analysis API → Dataset benchmark: runs benchmark
backend/app.py - Dataset benchmark → Chrominance DCT: analyzes image
backend/dataset_manager.py - Dataset benchmark → Micro-texture analysis: analyzes image
backend/dataset_manager.py - Dataset benchmark → Frequency analysis: analyzes image
backend/dataset_manager.py - Dataset benchmark → Multimodal fusion: fuses findings
backend/dataset_manager.py - CNN training → Dual-stream CNN: trains model
backend/models/trainer.py