How starkbyte-07/truevision works
TrueVision is a deepfake-forensics platform: users upload media through a web dashboard, which sends it to a local FastAPI proxy and then to an EC2 inference service. The documented assessment flow includes frame/face processing, tri-model analysis, calibrated verdicts, and forensic results; user feedback enters an active-learning loop. The graph preserves the documented email-report capability and training/calibration layer. Some server internals are unsampled, so those stage links rely on the README rather than inspected call sites.
Main entry points: ec2_server/app.py Inference API · Dashboard.jsx Dashboard workflow · backend/app.py FastAPI proxy · index.js API client
Components 21
Web application
- Dashboard workflow
truevision-app/src/pages/Dashboard.jsx - Upload zone
truevision-app/src/components/UploadZone.jsx - Processing view
truevision-app/src/components/ProcessingView.jsx - Results and feedback
truevision-app/src/components/ResultsPanel.jsx - Scan history
truevision-app/src/context/HistoryContext.jsx - Local user session
truevision-app/src/context/AuthContext.jsx - API client
truevision-app/src/api/index.js
Proxy and reporting
- FastAPI proxy
backend/app.py - Upload route
backend/routes/upload.py - Video frame extraction
backend/services/preprocessing.py - Email reports
backend/services/report.py
Forensic inference
- Inference API
ec2_server/app.py - Frame preprocessing
ec2_server/preprocessing.py - Face detection
ec2_server/facedetection.py - Tri-model inference
ec2_server/inference.py - Artifact analysis
ec2_server/artifact_analyzer.py - Confidence calibration
ec2_server/backend/calibration.py
Model learning
- Training and fine-tuning
ec2_server/backend/train.py - Feedback archive
ec2_server/app.py
Other components
- Media user
- SES / SMTP
Connections 23
- Media user → Dashboard workflow: submits media
README.md - Dashboard workflow → Upload zone: renders
truevision-app/src/pages/Dashboard.jsx - Dashboard workflow → API client: processes file
truevision-app/src/pages/Dashboard.jsx - Dashboard workflow → Processing view: shows progress
truevision-app/src/pages/Dashboard.jsx - Dashboard workflow → Results and feedback: shows results
truevision-app/src/pages/Dashboard.jsx - Dashboard workflow → Scan history: records scans
truevision-app/src/pages/Dashboard.jsx - Media user → Local user session: signs in locally
truevision-app/src/context/AuthContext.jsx - API client → FastAPI proxy: sends requests
README.md - FastAPI proxy → Inference API: forwards media
README.md - Upload route → Video frame extraction: extracts frames
backend/routes/upload.py - Inference API → Frame preprocessing: preprocesses media
README.md - Inference API → Face detection: detects faces
README.md - Face detection → Tri-model inference: provides face inputs
README.md - Tri-model inference → Artifact analysis: analyzes artifacts
README.md - Tri-model inference → Confidence calibration: calibrates scores
README.md - Inference API → Results and feedback: returns verdict
README.md - Results and feedback → API client: submits feedback
truevision-app/src/components/ResultsPanel.jsx - API client → Inference API: posts feedback
README.md - Inference API → Training and fine-tuning: fine-tunes ETCNN
ec2_server/app.py - Inference API → Feedback archive: archives predictions
ec2_server/app.py - Training and fine-tuning → Tri-model inference: updates ETCNN
README.md - FastAPI proxy → Email reports: generates reports
README.md - FastAPI proxy → SES / SMTP: sends email
README.md