Overview
I've wanted to work on my hardware and IoT skills for a while, so I've been continuing with Project Guardian!
The basic idea is I want to mount a camera to a persons shoulder. The camera needs to be able to rotate up, down, left and right. The camera will also have a laser attachment fixed to it so that this can be toggled.
Once those areas are implemented, we use object detection RNN's to get an objects position relative to the camera, and rotate to make the camera face the object.
Key processes here are;
- Precise servo control.
I have new servos that have sub-mm accuracy and can rotate to a specific point rather than just specifying direction.
- Chassis design.
Not my usual endeavour, I need a way to mount the camera and servo's so they can rotate on a hinge.
- Board communication.
Its unlikely the board can do the object tracking by itself, so I will likely end up streaming the video to a mobile device and returning the adjustments required.
- Recursive Neural Networks
Most of my previous experience is in decision based networks, CNN's usually. Having a network return not only a detection result, but a relative bounds position will be a good challenge
This project is serving as an opportunity for me to learn more about robotics and microprocessors.
My key learning outcomes here are:
Progress is ongoing with this project. At its current state I have an ESP32-DevKit microprocessor that has two servo's connected to it. The processor is powered by a standard battery pack, and the motors are powered by 4x AA batteries.
The basic idea is I want to mount a camera to a persons shoulder. The camera needs to be able to rotate up, down, left and right. The camera will also have a laser attachment fixed to it so that this can be toggled.
Once those areas are implemented, we use object detection RNN's to get an objects position relative to the camera, and rotate to make the camera face the object.
Key processes here are;
- Precise servo control.
I have new servos that have sub-mm accuracy and can rotate to a specific point rather than just specifying direction.
- Chassis design.
Not my usual endeavour, I need a way to mount the camera and servo's so they can rotate on a hinge.
- Board communication.
Its unlikely the board can do the object tracking by itself, so I will likely end up streaming the video to a mobile device and returning the adjustments required.
- Recursive Neural Networks
Most of my previous experience is in decision based networks, CNN's usually. Having a network return not only a detection result, but a relative bounds position will be a good challenge
This project is serving as an opportunity for me to learn more about robotics and microprocessors.
My key learning outcomes here are:
- Designing circuits to account for voltage restrictions
- The use of motors / servos in a precise fashion
- Using a microprocessor to control camera, servos and a wifi stream to connect this system to a phone app
- Using ML tech for basic facial recognition
Progress is ongoing with this project. At its current state I have an ESP32-DevKit microprocessor that has two servo's connected to it. The processor is powered by a standard battery pack, and the motors are powered by 4x AA batteries.