DepthAI ML Training
This repository contains colab notebooks that provide step by step tutorials on how to:
- Train your own ML model - either on default or custom dataset
- Convert the ML model so it's compatible with the DepthAI platform - conversion docs here
- Deploy the ML model to the OAK camera, so it runs on the accelerated hardware on the edge
Check Out RoboFlow!
If you are new to the ML world, we suggest using RoboFlow for annotating, training and deploying ML models, especially if you are new to the ML world. They have created a course for the DepthAI platform that is a great starting point. They also have two blog posts, quickstart guide and training custom object detection model.
What is DepthAI?
is the platform for Spatial AI. 4 TOPS on a tiny device which can ouput structured data of what is in the physical world - what it is, and where it is in meters (XYZ) - in real-time.
What is Google Colaboratory?
Google Colaboratory allows you to train neural models yourself using their fast GPU instances, and in some cases (depending on the dependencies of the training frameworks), even using the Google TPU - all for free!
It is exactly intended for this proof of concept and initial research. And if you hit the limits of the free account, you can upgrade to a Pro version of Google Colab for only $10/month.