glhr/color-classifier

Classify the color of an object in an image using machine learning

Color-classifier

Implementation of an object color classifier using machine learning and a small custom dataset. This was developed for the Little Helper 7+ at Aalborg University, however it can also be run as a stand-alone application.

This was developed using classifiers which support online learning in scikit-learn. For feature description, histograms in different colour spaces are compared.

Requirements

pip3 install -r requirements --user

The utils package here is also required:https://github.com/glhr/img-utils

Project structure

├── classifier.py      # main file, for classifying objects
├── classifierutils.py # for loading classifier parameters and models, loading/saving the dataset
├── dataset_img        # contains the images used to train the original model
├── dataset_json       # dataset for different colour spaces and histogram resolutions
├── dataset_plots      # images generated for documentation
├── dataset.py         # for generating a JSON dataset file from the training images
├── dataset_user       # images gathered through incremental learning
├── eval.py            # for comparing the performance of different classifiers and parameters
├── requirements.txt
├── test               # images used and generated during testing
└── test.py            # for testing incremental learning