This project offers a LAN-based AI monitoring framework that repurposes Android smartphones as network cameras for local data acquisition and analysis. It provides real-time video streaming and AI-driven event detection with structured output, targeting users needing a privacy-focused, cost-effective, and extensible visual intelligence prototype.

How It Works

Sentinel uses a "mobile capture + PC processing + browser control" architecture. An Android app (CamFlow) uploads JPEG frames via HTTP POST to a PC Flask server. The server streams MJPEG video, handles recording, and integrates an optional AI module. This AI component employs a cost-efficient, layered approach combining traditional CV for triggers with model inference for semantic analysis, using a state machine to optimize AI calls and generate structured JSON outputs.

Quick Start & Requirements

• Installation: Clone repo, set up Python virtual environment, pip install -r requirements.txt. Install the CamFlow-v1.0.0-beta.apk on Android or build from source.
• Prerequisites: PC: Python 3.9+ (Win/macOS/Linux), Git. Android: 8.0+. Both devices must be on the same LAN. Key Python dependencies: Flask, NumPy, OpenCV.
• Links: Repo: https://github.com/suzuran0y/CCTV-Smartphone-AI-Monitoring.

Highlighted Details

• Smartphone Repurposing: Uses existing Android devices as network cameras, cutting hardware costs.
• Browser Preview: Real-time MJPEG video streaming viewable directly in web browsers.
• Structured AI Output: AI analysis yields controllable JSON for events, risk levels, and summaries, aiding data analysis.
• Privacy-Focused: Operates entirely on a local LAN, keeping data local and avoiding cloud dependencies.
• Efficient AI Triggering: Layered "CV + model inference" with a state machine optimizes AI calls and cost.
• Auto Discovery: UDP-based server discovery simplifies client connection.

Maintenance & Community

The README provides no specific details on active maintenance, contributors, community channels, or a public roadmap beyond planned future improvements. The last update mentioned is February 26, 2026.

Licensing & Compatibility

Released under the MIT License. Suitable for learning and research; production deployment requires implementing additional security (authentication, encryption) and privacy measures.

Limitations & Caveats

The project is in beta (v1.0.0-beta). Production use demands significant work on security, privacy, and stability. AI integration is primarily for Volcengine; other models require code modification. Android client requires enabling installation from unknown sources.