| --- |
| sdk: gradio |
| title: 🔬 Forensic Triage & Postmortem Intelligence System |
| emoji: 🔬 |
| colorFrom: red |
| colorTo: gray |
| sdk_version: 6.14.0 |
| python_version: '3.13' |
| app_file: app.py |
| pinned: false |
| tags: |
| - ml-intern |
| - forensics |
| - nlp |
| - investigation |
| - medical |
| short_description: AI-powered forensic investigation support system |
| --- |
| |
| # 🔬 AI-Powered Forensic Triage & Postmortem Intelligence System |
|
|
| An intelligent investigative support system that integrates AI, NLP, and data correlation techniques to assist forensic investigations. |
|
|
| ## Key Features |
|
|
| | Module | Description | |
| |--------|-------------| |
| | 📄 **Autopsy Report NLP** | Extracts forensic entities (cause of death, injuries, toxicology) from unstructured reports | |
| | ⏱️ **Time-of-Death Estimation** | Henssge nomogram + postmortem indicators for PMI calculation | |
| | 📱 **Digital Evidence Correlation** | Correlates CCTV, mobile metadata, and geolocation data | |
| | ⚠️ **Risk Scoring & Anomaly Detection** | Multi-factor risk assessment with pattern identification | |
| | 📅 **Investigation Timeline** | Integrated visualization combining all evidence sources | |
|
|
| ## How to Use |
|
|
| 1. **Autopsy Report**: Upload or paste an autopsy report → system extracts key forensic entities |
| 2. **Time of Death**: Enter body temperature and postmortem signs → get PMI estimate with cooling curve |
| 3. **Digital Evidence**: Upload CCTV/mobile logs as CSV → correlations and patterns identified |
| 4. **Risk Score**: After analyzing evidence, compute multi-factor risk score |
| 5. **Timeline**: Build an integrated timeline from all evidence sources |
|
|
| ## Methodology |
|
|
| - **Henssge Nomogram (1988)**: Double-exponential body cooling model for PMI |
| - **Pattern-based NLP**: Forensic-domain entity recognition with confidence scoring |
| - **Multi-factor Risk Engine**: Weighted scoring across violence, toxicology, digital evidence, temporal consistency |
| - **Anomaly Detection**: Identifies cross-factor inconsistencies (e.g., defensive wounds without homicide classification) |
|
|
| ## ⚠️ Disclaimer |
|
|
| This system is **strictly an investigative assistance platform**. It is NOT a replacement for forensic experts, medical professionals, or legal authorities. All outputs are intended to support human decision-making, not automate legal conclusions. |
|
|
| ## Technical Stack |
|
|
| - **Frontend**: Gradio |
| - **NLP**: Regex-based forensic entity extraction |
| - **Math**: Scipy (Henssge solver), NumPy |
| - **Visualization**: Plotly (interactive charts, timelines, radar plots) |
| - **Data**: Pandas for evidence processing |
|
|
