Android Smartphone Based Visual Object Recognition for Visually Impaired Using Deep Learning

Abstract

In this paper, we propose a deep learning based visual object recognition model, to guide a visually impaired person in an outdoor environment using a smartphone equipped with internet connectivity. The deep learning based computing is performed by the high-end server and is to be hosted on a cloud. The boosted optimized convolutional neural network (CNN) based visual object recognition model is proposed by experimenting various CNN models, different optimizers, and varying learning rate. Visually impaired person captures the video and sends frame as input to the server for object segmentation and object recognition using proposed model. He then receives the information about surrounding objects and obstacles as voice input to the hearing device attached to the smartphone. The proposed system provides measurable high accuracy for recognition of 11 outdoor objects, to guide visually impaired person more effectively than traditional guiding cane.