ML Tutorial - Offset-official/Nautilus-One GitHub Wiki

ML Tutorial

This is the documentation of the Machine Learning methods used in the project.

Object Detection

We used YOLO models for object detection on the Jetson Nano. Follow the steps:

1. Collect dataset for objects of interest (simulation/real-world).
2. Annotate your dataset.
3. Train your YOLO model. We suggest using "s" model and exporting them to TensorRT (.engine) format for fast inference on the Nano.
4. You can use ultralytics Python library for streamlined training, export, and inference. Our codebase has the corresponding code.

Reference for training: Kaggle Notebook