AI ML Robotics

Raspberry Pi is a little wonder capable of running variety of Machine Learning models on it. Also, plenty of pre-trained Object Detection & Image Classification ML models are available on TensorFlow website free of cost. So why not exploit these resources to build an advanced robot for ourselves. Let's push it to limits by attaching a Coral USB Accelerator too.

Replicate Robot Software: Install the complete Earthrover Robot software on your Raspberry Pi using the bash script introduced in this article. The bash script will install a Webserver, Tensorflow Lite & Coral Coral USB Accelerator Libraries, OpenCV and Robot software with AI functions such as Object Detection / Image classification with relevant ML models on your Raspberry Pi. All you need is a Raspberry Pi and a Picamera to embark this learning journey.

Human Following Robot: The robot uses Object Detection ML model to detect a person in the camera frame. The model provides coordinate information of the person in the frame. The fun part is to write a Python code to use this information and move the robot in order to bring the person in the center of the frame. What you get as a result is a Human Following Robot. You will be amazed and satisfied when your robot follows you like a loving pet.

Object Tracking Robot: Learn how to use an Object Detection Model on a Raspberry Pi to track a particular object of interest. Monitor the Robot's eye view on a browser and the whole tracking process in action by streaming the camera frame over the LAN using OpenCV and Flask. Learn to use Google Coral Accelerator to make the inferencing fast enough for Tracking requirements.

Object Detection by Robot: The robot makes use of a pre-trained Object Detection Machine Learning model to identify objects around it. It can receive commands through a Web GUI to check for a specific object in the camera frame. When the object is detected an alert is generated on the Web GUI. You can tweak the code to scare off dogs or cats and welcome a person using this robot. You will learn how to implement TensorFlow Lite, FLASK and OpenCV in the process.

Teachable Machine on Robot: Teachable Machine online tool by Google is an excellent Web based tool to create ML models rapidly. The project teaches you how to use this tool to create a custom model, download and export it on a raspberry pi and tweak the sample javascript code to meet your purpose. Ultimately, building a gesture controlled robot.