Open AccessObject recognition using cognitive artificial intelligence’s knowledge growing system: A preliminary study
Author(s) -
Arwin Datumaya Wahyudi Sumari,
Ika Noer Syamsiana,
Dimas Rossiawan Hendra Putra,
Rosa Andrie Asmara
Publication year - 2021
Publication title -
iop conference series. materials science and engineering
Language(s) - English
Resource type - Journals
eISSN - 1757-899X
pISSN - 1757-8981
DOI - 10.1088/1757-899x/1073/1/012071
Subject(s) - artificial intelligence , computer science , machine learning , object (grammar) , artificial neural network , benchmark (surveying) , geodesy , geography
Object recognition has been a challenge for an intelligent system. There have been various approaches to develop such a system by utilizing machine learning especially which are based on neuron that is, neural network and deep learning. Common problems when using those approaches are the first one is dataset availability and the second one is the number of data. Lack of data causes neural-based approaches cannot be well operated, while a small number of data causes low accuracy results on the system. From another point of view, a considered-new technology from Cognitive Artificial Intelligence (CAI) perspective called as Knowledge Growing System (KGS) which may cope with such problems. With the capability to build its own knowledge from nothing, KGS is able to carry out recognition while developing its knowledge regarding the phenomenon it is trying to recognize. In this research, we showed KGS capability to perform object recognition as it is developing knowledge when interacting with such object directly. We did a benchmark on face recognition use-case with some common machine learning methods to show their performance on a small number of data, and KGS showed good results. With 100 feature-set from 5 persons’ face images, KGS achieves Degree of Certainty (DoC) as much as 80% which is the system’s prediction accuracy that enables it recognizing the person based on that-moment data. Even though it is still lower compared to machine learning methods, but KGS shows advantages that it does not require high computational cost because it requires no training and no model development. In this research, we also showed that KGS enables the fast-deployment light-operated object recognition system.