Howdy!
Through weeks of R&D, trial and error, and a lot of summer time on my sleeves, I would like to release a document that attempts to tackle and raise the barrier of the V5 Brain’s functionality. Additionally, the plan is to make the barrier-of-entry be as streamlined as possible for even novice individuals. Therefore, through my research I have created instructions on how to create a computer system in Ubuntu that has the following capabilities:
- CUDA Computation with a graphics card
- cuDNN that optimizes neural networks with CUDA
- Docker
- CUDA Container to work with Docker
- Installing Miniconda
- Installing ROS, or Robot Operating System
- Realsense SDK for a Realsense D435i Camera
- Calibrating the D435i Camera’s IMU
- Calibrating the D435i Camera’s Depth Sensing
- ORB_SLAM3 for Simultaneous Localization and Mapping with the Realsense D435i Camera
- Reading data from ORB_SLAM3 to get x, y, z, pitch, yaw, roll data from the camera via a Python script reading the ROS output stream.
- OpenCV for Computer Vision
- PyTorch for YOLOv5 functionality and Robot Recognition
- Using a Logitech C920 Webcam, and tuning the focus on it, for Object Detection using PyTorch
- A simple communication method by building my own send/receieve functions both with and from the V5 Brain with a mere USB cable
It is important to note that ORB_SLAM3, and any other SLAM method, requires loads of calibration. It may not work the first time and you will need to do a lot of tuning. Good luck!
