Computer Vision Rock Paper Scissors

Requirements

Python 3
A webcam

Installation

We recommend using a Python virtual environment.

# Build the virtual environment
python3 -m venv rps-env

# Activate the virtual environment
source rps-env/bin/activate

To install Python dependancies run the following

pip install -r requirements.txt

If Kivy fails to install, please consult platform specific install instructions here https://kivy.org/doc/stable/gettingstarted/installation.html

The application can now be started by running

python main.py

Configuration

Several configuration values are read in from a config.json file. If config.json does not exist default values will be used. To change any values copy config.json.example to config.json

confThreshold: Object detection confidence threshold from 0.0 to 1.0.
nmsThreshold: Non-maximum suppression threshold value
wait_interval: Wait time in seconds between Rock, Paper, Scissors, Shoot
corrections_path: Folder path to save mis-detected images in

Labelbox

Uploading images with misdetected gestures to Labelbox is supported. A Labelbox API key and dataset ID can be set in config.json. API keys can be generated under the Account section of your Labelbox account. Dataset IDs can be found in Labelbox URLs. Click on a dataset from the Datasets tab and then copy the string after dataset/ in the URL. For example, copy x's from https://app.labelbox.com/dataset/xxxxxxxxxxxxxxxxxxxxxxxxx

Tests

The pytest framework is used to preform unit tests. From the root project directory in the terminal, run

pytest