On one approach to recognizing fuzzy images of faces based on an ensemble

This article implements and analyzes face recognition algorithms based on fuzzy image representations using several main classifiers, including Decision Tree, Random Forest, and K-Nearest Neighbors. The main focus is on the extraction of informative features to correct recognition and classification of the human face. The algorithms under study recognize a person by their eyes if their face is covered with something. According to the research results, the accuracy of the KNN, Random Forest and Decision Tree algorithms is 95.5%, 96.1% and 92.3%, respectively. In order to improve the results, the ensemble learning method was implemented, which allows a combination of several classification results using a metaclassifier. The first level of ensembling uses the three basic models mentioned above to predict the output of the test datasets. In the second level of this process, a new result was generated, taking all the results of the underlying models as input. As a result of the ensembling, the accuracy was 96.8%, exceeding the accuracy of each algorithm. A comparative assessment of the developed ensemble algorithm quality with other ensemble algorithms based on the boosting approach is also made.