Automobile Parts Localization Using Multi-layer Multi-model Images Classifier Ensemble

Abstract

To help in the education-assistive technology in automobile industry, one of the best approaches is to create an augmented reality of each car part in the engine room. To do so, the location of each part is required to be identified. In this paper, the Multi-layer Multi-model Images Classifier Ensemble (MICE) is utilized in the recognition process. The data was collected from car engine room of the Toyota Vios 2017 model in different angles and different lighting conditions. The recognition rate of the best validation test set is 91.25% even though there is still problem with the similarity between classes.