Fusion based En-FEC Transfer Learning Approach for Automobile Parts Recognition System

Abstract

The artificially supervised classification of real world entities have gained a phenomenal significance in recent year of computational advancements. An intelligent classification model focuses on rendering accurate outcomes vide the implicated paradigms with respect to the subjected data employed to train the classifier. This paper proposes a novel deep learning approach to classify the various parts of any operational engine such as crank shafts, rock-arms, distributer, air duct, assecorybelt etc. deployed in automobiles. The proposed architecture distinctively utilizes convolution neural networks for this typical classification problem and altogether constructs a robust transfer learning paradigm to render the correct class label against the validation and test images as the conclusive result of the classification. The proposed methodology poses in such a way that it can qualitatively classify and henceforth give the corresponding class label of the machinery/engine part under consideration. This computationally intelligent architecture requires the user to feed the image of the engine part to the model in order to achieve the requisite responses of classification. The main contribution of the proposed method is the development of a robust algorithm that can exhibit pronounced results without training the entire ConvNet architecture from scratch, thereby enabling the proposed paradigm to be deployable in application instances wherein limited labeled training data is available.