How shubhigupta155/verimediaai works
VeriMedia AI is a digital-media forensics application: users submit media, specialized image, video, audio and lip-sync analysis produces evidence, and fusion estimates authenticity risk. The Streamlit excerpt shows a working image-analysis path; the FastAPI route currently validates uploads but explicitly says detector orchestration is not connected. Other domain stages and explanation/report capabilities are documented in the README, but their runtime wiring is not established by the excerpts. Edges for those documented stages cite the README; unsampled implementation details are not inferred.
Main entry points: image/pipeline.py Image forensics · fusion/pipeline.py Evidence fusion · audio/pipeline.py Audio forensics · video/pipeline.py Video forensics
Components 25
User Interfaces
- Streamlit interface
app.py - FastAPI service
backend/main.py - Upload validation
backend/routes/analysis.py
Forensic Analysis
- Image forensics
modules/image/pipeline.py - Metadata analysis
modules/image/metadata.py - Compression analysis
modules/image/compression.py - Noise analysis
modules/image/noise.py - Frequency analysis
modules/image/frequency.py - AI image detector
modules/image/detector.py - Image heatmap
modules/image/explainability.py - Video forensics
modules/video/pipeline.py - Deepfake detector
modules/video/deepfake_detector.py - Xception model
modules/video/network/xception.py - Audio forensics
modules/audio/pipeline.py - Audio preprocessing
modules/audio/preprocessing.py - Spectrogram analysis
modules/audio/spectrogram.py - Voice manipulation detection
modules/audio/voice_detector.py - Lip-sync analysis
modules/lipsync/pipeline.py
Evidence Assessment
- Evidence fusion
modules/fusion/pipeline.py - Result validation
modules/fusion/validator.py - Fusion strategies
modules/fusion/strategies.py - Risk assessment
modules/fusion/risk_engine.py
Explanation & Results
- DeepSeek explanation
ai/deepseek.py - Forensic reporting
modules/report.py
Other components
- Media user
Connections 26
- Media user → Streamlit interface: submits media
README.md - Streamlit interface → Image forensics: runs analysis
app.py - Media user → FastAPI service: uploads media
README.md - FastAPI service → Upload validation: registers route
backend/main.py - Upload validation → Media user: accepts upload
backend/routes/analysis.py - Image forensics → Metadata analysis: extracts metadata
modules/image/pipeline.py - Image forensics → Compression analysis: analyzes compression
modules/image/pipeline.py - Image forensics → Noise analysis: analyzes noise
modules/image/pipeline.py - Image forensics → Frequency analysis: analyzes frequencies
modules/image/pipeline.py - Image forensics → AI image detector: detects generation
modules/image/pipeline.py - Image forensics → Image heatmap: generates heatmap
modules/image/pipeline.py - Video forensics → Deepfake detector: analyzes deepfakes
- Deepfake detector → Xception model: uses model
modules/video/deepfake_detector.py - Audio forensics → Audio preprocessing: preprocesses audio
modules/audio/pipeline.py - Audio forensics → Spectrogram analysis: generates spectrogram
modules/audio/pipeline.py - Audio forensics → Voice manipulation detection: detects manipulation
modules/audio/pipeline.py - Video forensics → Lip-sync analysis: supports sync analysis
README.md - Image forensics → Evidence fusion: provides evidence
README.md - Video forensics → Evidence fusion: provides evidence
README.md - Audio forensics → Evidence fusion: provides evidence
README.md - Lip-sync analysis → Evidence fusion: provides evidence
README.md - Evidence fusion → Result validation: validates results
modules/fusion/pipeline.py - Evidence fusion → Fusion strategies: fuses scores
modules/fusion/pipeline.py - Evidence fusion → Risk assessment: assesses risk
modules/fusion/pipeline.py - Evidence fusion → DeepSeek explanation: supplies evidence
README.md - DeepSeek explanation → Forensic reporting: explains findings
README.md