jiteshsaini / robotics-level-4

This repo contains projects created using TensorFlow-Lite on Raspberry Pi and Teachable Machine. AI and ML capabilities have been integrated with Robot's software.

Visit Website:

Youtube Channel:

Robotics Level 4

This repo is an extension of previous level. The code of this robot is organised in various folders inside the directory 'earthrover'. The names of these folders briefly indicate the purpose of the code inside them. This repo focusses on the advanced capabilities embedded into the robot via use of Pre-trained Machine Learning models provided by "tensorflow.org" or created via online tool of Google called Teachable Machine. The following projects in this repo demonstrate how we can integrate Tensorflow Lite and such Machine Learning Models on a Raspberry Pi computer. You can further read about them by accessing their individual README.md file.

• Gesture Controls
• Image Classification
• Object Detection
• Object Tracking
• Human Following

Download the code and configure your Raspberry Pi

I have created a bash script that installs all the packages / libraries required to run this code on your Raspberry Pi. The script also downloads the code of this repo along with ML models on your device automatically. Follow the instructions on the link given below to configure your Raspberry Pi:-

https://helloworld.co.in/earthrover

Object Detection

The code for this project is placed in a directory named 'object_detection' inside the 'earthrover' directory The ML model used in this project is placed inside 'all_models' directory.

The robot can spy on a particular object and provide an alarm on a remote Web Control panel whenever the selected object appears in the frame.

Object Tracking

The code for this project is placed in a directory named 'object_tracking' inside the 'earthrover' directory The ML model used in this project is placed inside 'all_models' directory.

Robot is made to track a ball and follow it. You can see the robot's camera view on a browser while it is tracking the ball.

Human Following

The code for this project is placed in a directory named 'human_following' inside the 'earthrover' directory The ML model used in this project is placed inside 'all_models' directory.

Robot is made to follow a human. It is a good human follower :)

Image Classification

The code for this project is placed in a directory named 'image_classification' inside the 'earthrover' directory. The ML model used in this project is placed inside 'all_models' directory.

The robot's camera view is streamed over LAN with overlays of image classification output. Also, if an object is recognised, the robot speaks out its name.

Gesture Control

The code for this project is placed in a folder named 'tm' inside the 'earthrover' directory. The model used in this project is trained through Teachable Machine online tool by Google. The model files are present in the same directory. Presently the model is trained to recognise hand gestures. You can train your own model using Teachable Machine and replace the model files to customise the project.

robotics-level-4's Issues

How do I adjust the GPIO's?

Where in the python code can I adjust the GPIO assignments for the motors?

Optimization of the main cycle

Remove from the main cycle of the program the addition of labels on the video and sending the frame to the browser. since this slows down the robot's reaction speed very much

Possible to integrate pan tilt for camera?

Would it be possible to change the UI to reflect a pan tilt mechanism like the one below please, so basic UP, DOWN, LEFT and RIGHT buttons like the Rover control buttons for the Camera?

low fps raspberry 3

I am trying to run object detection on raspberry pi3. program runs successfully without any errors but the fps is as low as 0.7.
how do i increase fps at least to 4 without coral accelerator ?

Webcam not opening in local host