Node.js app for Tactode without ArUco markers

Computer vision algorithm developed in JavaScript/Node.js that performs segmentation and 2D recognition of Tactode tiles in an image. This work was developed during a M.Sc. thesis named "Computer Vision System for Tactode Programming". The document is available here.

Necessary tools

Ubuntu

• Git: sudo apt install git
• npm: sudo apt install npm
• node sudo apt install nodejs

Installation

Start by cloning the repo: git clone https://github.com/daniqsilva25/tactode_minus_ArUco_App

Install app dependencies: cd tactode_minus_ArUco_App && npm install

Scripts explained

• tactode_cv_alg.js: this is the main script containing the computer vision algorithm for segmenting and classifying tiles
• params.js: this is an auxiliary script containing global classes and functions
• usage.js: this is a script that exemplifies how to use the main script and its returning values

Folders explained

• /debug: folder contaning the debug images
• /test_imgs: folder containing the test images
• /trained_models: folder containing configurations of HOGs, trained SVM models and a .txt file containing the definition of the tiles classes

Usage

Running pre-defined scripts

Run a pre-defined test script without debug (does not save the resulting image): npm test

Run a pre-defined test script with debug (saves the resulting image): npm debug

Run a pre-defined JavaScript standard verification script: npm verify-std

Running custom scripts

node usage.js [image] [debug]

Arguments explanation

• image: image path for running the algorithm
• debug: with debug ('debugOn') or without debug (empty space)

Examples

• node usage.js test_imgs/img_6.jpg
• node usage.js test_imgs/img_3.jpg debugOn