Restructure the project to run locally on Raspberry Pi with computationally-intensive steps offloaded to another device on the same network

[eskutcheon]

Due to the planned obsolescence of the Jetson Nano line and the issues with using its GPIO headers, I've decided to do a major refactor to run the system on a Raspberry Pi while offloading any computationally-intensive parts (primarily everything involving object detection) on my desktop in a container. It's still being planned, so things are subject to change, but this will be done in a way to minimize additional costs and to keep it maintained primarily for my own personal use.

Currently, the refactor is planned to be something like

Purr-Patrol-Turret/
├── src/
│ ├── local/ # Code running on the Raspberry Pi
│ │ ├── controller.py # Manages modes and GPIO operations
│ │ ├── motion_tracking.py # Handles camera feed and frame capture
│ │ ├── turret.py # Turret control (GPIO and motor logic)
│ │ ├── server_messaging.py # NEW: Handles communication with desktop
│ │ └── requirements_pi.txt # Dependencies specific to the Raspberry Pi
│
│ ├── remote/ # Code running on the desktop
│ │ ├── server.py # NEW: Flask server to process frames
│ │ ├── object_detection.py # Detection logic with GPU acceleration
│ │ ├── requirements_desktop.txt # Dependencies specific to the desktop
│
│ ├── shared/ # Code common to both
│ │ ├── utils.py # Shared utilities for both systems
│ │ ├── config.py # Configuration (shared settings)

[eskutcheon]

UPDATE:

I'm planning to simplify this with a direct LAN connection between my desktop running the more intense image processing and instance segmentation code and the Raspberry Pi operating the robot.

This would greatly reduce the communication overhead, simplify the code quite a bit, reduce the attack vector surface, and avoid a lot more configuration of the Pi, which wouldn't need to be connected to any network at that point.

Security

To further ensure better security, I'm considering including a shared secret key in HTTP headers or gRPC metadata to validate requests. If I want to use encrypted channels, I'd generate a self-signed SSL certificate for the desktop and configure Flask (server.py) to use the certificate.