WalkSafe

Urban navigation system combining street-level safety modeling with graph-based routing.

WalkSafe computes safer pedestrian routes by integrating machine-learned safety scores into a shortest-path search over real city street graphs.
The system combines data on crime density, lighting, and proximity to emergency services to estimate relative safety at the street level.

---

🧠 Overview

WalkSafe models urban safety as a spatially varying risk field.
Each street segment (graph node) is assigned a safety score predicted by a Random Forest trained on real geospatial data.
Routing then minimizes a weighted cost function:

Cost Function:
cost(edge) = distance + λ × (1 - safety)

By adjusting λ, users can explore the trade-off between the shortest and the safest path.

---

🧩 System Architecture

                   ┌────────────────────────────┐
                   │   Raw Geospatial Datasets   │
                   │ (Crime, Lighting, Services) │
                   └──────────────┬──────────────┘
                                  │
                                  ▼
                    ┌────────────────────────────┐
                    │   Feature Extraction Layer  │
                    │  (GeoPandas + OSM data)     │
                    └──────────────┬──────────────┘
                                  │
                                  ▼
                   ┌──────────────────────────────┐
                   │  Random Forest Safety Model   │
                   │ (scikit-learn, trained offline)│
                   └──────────────┬───────────────┘
                                  │
                                  ▼
                   ┌──────────────────────────────┐
                   │   Precomputed Node Scoring    │
                   │ (Applied to OSM street graph) │
                   └──────────────┬───────────────┘
                                  │
                                  ▼
                   ┌──────────────────────────────┐
                   │  Routing Engine (Flask API)   │
                   │  Dijkstra search w/ λ-safety  │
                   └──────────────────────────────┘

---

🧰 Tech Stack

• Python, Flask, React
• scikit-learn, GeoPandas, osmnx, Folium
• Docker for containerization
• GitHub Actions for continuous integration and testing

---

⚙️ Development Setup

Clone and install dependencies:

git clone https://github.com/rahulgh33/WalkSafe.git
cd WalkSafe
pip install -r requirements.txt

Run locally:

python -m src.api.app

Containerize (optional):

docker build -t walksafe .
docker run -p 5001:5000 walksafe

---

📖 Description

WalkSafe is designed as an experimental framework for data-driven urban navigation, emphasizing explainability and public-safety modeling.
Its modular design allows substitution of models, additional features (e.g., temporal crime variation), or front-end map layers for real-time interaction.