Abstract
Indoor object classification is a key element for indoor navigation assistance systems. Indoor objects knowledge helps Visually Impaired People (VIP) in their indoor navigation and facilitates their daily life. This paper proposes a new classification system used especially for indoor object recognition based on Deep Convolutional Neural Network (DCNN) model which can be implemented on mobile embedded platforms.
Experimental results obtained using natural images (with natural illumination) from the MCIndoor 20000 dataset show that the proposed approach achieves almost100% accuracy for indoor object classification.