AiCore-Computer-Vision-Project

This project uses computer vision to implement a new version of the classic Rock Paper Scissors(RPS) game. In this version, the user will use his/hers computer/laptop camera to play the game with the computer as the opponent.

Milestone 1

As a first milestone for this project, I have used Teachable Machines, a web-based tool for machine learning models. I have trained a model with images of myself, such that the model can recognise four different image categories: Rock, Paper, Scissors or Nothing. Then, I have downloaded this trained model to my local machine.

Milestone 2

For the next part I have created a conda environment for my project and installed the necessary libraries and dependencies:

• opencv python - a library for computer vision
• tenserflow - a library for machine learning & AI
• ipykernel - a package that provides a IPython kernel for Jupyter

Milestone 3

I began implementing the computer vision version of RPS by defining a class called RPS_game(). The class had several methods that represent the backbone of the project, namely:

• methods that signal the start and stop of frame acquisition for the camera
• methods that record the computer's choice and the user's choice by incorporating feedback from the camera and the trained model
• a method for determining the winner for each round of the game

Milestone 4

Finally, I have defined a function play_game() that incorporates methods from the RPS_game class. In addition, it introduces a new functionality that display a countdown timer (3, 2, 1, GO!) at the start of each round. The user is advised to show his/her choice after the GO! cue. Inside this function, the winner is announced after a 5 rounds game.

Conclusion

I have built a computer vision project, that makes the classical RSP a bit more fun! Besides the functionalities already added, future versions of this game could benefit from additional improvements such as:

• adding a textbox that displays round number
• adding a textbox that display the winner for a specific round
• adding a functionality that allows the user to press 'next' (on the camera display) for a new round to begin

Video 1: Video showing an outline of the computer vision project

project-outline_us6MHamo.mp4

Video 2: Video showing the training classes & model performance on the Teachable Machine website

machine-learning-model_N5PpvTIu.mp4

Video 3: Video showing an example RPS game

play-game-example_9eZZ4LzW.copy.2.mp4