Hardware Notes

Rough notes on Hardware, advice, questions, ideas etc.

This was a project we worked on over the course for 3 months getting assitance from UW Harvard and MIT professors. This is a resource area.

Process

Select a feature below to learn more about it.

Model Diagram

CAD Model Digrams

View

Model Render

CAD Model Render.

View

Questions and Advice for SentryAI Device

Questions

  • What if someone jumps through windows?
  • Blurriness, covering hoods, image technicalities?
  • Potential errors through USB camera and mic connection?
  • Image compression concerns.
  • Identifying sounds similar to gunshots.
  • USB C range extender feasibility?
  • Power over Ethernet capabilities?

Advice

  • Implement the system throughout the entire building, not just in specific areas.
  • Focus on both facial and threat recognition.
  • Ensure resiliency through minimal moving parts.
  • Be cautious of potential errors through USB connections.
  • Prioritize obtaining raw, uncompressed images.
  • Use Raspberry PI supported camera modules.
  • Employ a trained model for threat detection.
  • Do not store any data, ensuring privacy.
  • Dispose of audio capture after a certain time (e.g., a day).
  • Avoid being an attendance tracker; names are not necessary, just images.
  • Update the model annually using reference images without storing them.
  • Ensure privacy by design, not just by default.
  • Consider Air Gapping Concept for enhanced security.
  • Implement License Plate Recorders (LPRs).
  • Explore day and night timer versions for cameras (infrared and regular).
  • Use 2FA with Bluetooth and Wifi alongside facial recognition.

Design

  • Use templates from Amazon for fake security camera modules to save costs.

Microphone

  • Avoid obstructions near microphones.
  • Use condenser mics for long-range audio capture.
  • Include two microphones, each with a 180-degree range.
  • Capture raw audio data without sound filters.
  • Ensure high fidelity and bitrate for audio capture.
  • Utilize stereoscopic mics for directional audio capture.
  • Reference "ShotSpotter" published paper for insights.
  • Distinguish between various sounds, including gunshots and similar noises.

Camera

  • Consider using ESP32CAM and USB repeater 4k webcams.
  • Reference Raspberry Pi Documentation for camera options.
  • Explore various camera models available on Amazon.

Motherboard

  • Options include NanoITX, NUKs Intel AMD, and Nanocontroller.
  • Consider using USB Accelerator | Coral for machine learning acceleration.
  • Explore Raspberry Pi 4 Model B options.

Miscellaneous

  • Ensure a robust battery backup system.
  • Maintain a constant Ethernet connection with backup.
  • Emphasize security and edge network connectivity.