scholarly journals An assistive model of obstacle detection based on deep learning: YOLOv3 for visually impaired people

2021 ◽  
Vol 11 (4) ◽  
pp. 3434
Author(s):  
Nachirat Rachburee ◽  
Wattana Punlumjeak
Keyword(s):  
Deep Learning ◽  
Visually Impaired ◽  
High Speed ◽  
Obstacle Detection ◽  
World Health ◽  
Experimental Result ◽  
Detection Accuracy ◽  
Visually Impaired People ◽  
Impaired People ◽  
Audio Output

<span>The World Health Organization (WHO) reported in 2019 that at least 2.2 billion people were visual-impairment or blindness. The main problem of living for visually impaired people have been facing difficulties in moving even indoor or outdoor situations. Therefore, their lives are not safe and harmful. In this paper, we propose</span><span>d</span><span> an assistive application model based on deep learning: YOLOv3 with a Darknet-53 base network for visually impaired people on a smartphone. The Pascal VOC2007 and Pascal VOC2012 were used for the training set and used Pascal VOC2007 test set for validation. The assistive model was installed on a smartphone with an eSpeak synthesizer which generates the audio output to the user. The experimental result showed a high speed and also high detection accuracy. The proposed application with the help of technology will be an effective way to assist visually impaired people to interact with the surrounding environment in their daily life.</span>

scholarly journals Object Identification and Safe Route Recommendation Based on Human Flow for the Visually Impaired

Sensors ◽  
2019 ◽  
Vol 19 (24) ◽  
pp. 5343 ◽  
Author(s):  
Yusuke Kajiwara ◽  
Haruhiko Kimura
Keyword(s):  
Visually Impaired ◽  
Object Identification ◽  
Obstacle Detection ◽  
World Health ◽  
Detection Methods ◽  
Monocular Camera ◽  
Visually Impaired People ◽  
Average Accuracy ◽  
Impaired People ◽  
Health Organization

It is difficult for visually impaired people to move indoors and outdoors. In 2018, world health organization (WHO) reported that there were about 253 million people around the world who were moderately visually impaired in distance vision. A navigation system that combines positioning and obstacle detection has been actively researched and developed. However, when these obstacle detection methods are used in high-traffic passages, since many pedestrians cause an occlusion problem that obstructs the shape and color of obstacles, these obstacle detection methods significantly decrease in accuracy. To solve this problem, we developed an application “Follow me!”. The application recommends a safe route by machine learning the gait and walking route of many pedestrians obtained from the monocular camera images of a smartphone. As a result of the experiment, pedestrians walking in the same direction as visually impaired people, oncoming pedestrians, and steps were identified with an average accuracy of 0.92 based on the gait and walking route of pedestrians acquired from monocular camera images. Furthermore, the results of the recommended safe route based on the identification results showed that the visually impaired people were guided to a safe route with 100% accuracy. In addition, visually impaired people avoided obstacles that had to be detoured during construction and signage by walking along the recommended route.

scholarly journals A Hands-Free Obstacle Detection and Avoidance System

2015 ◽  
Vol 5 (3) ◽  
pp. 801-804
Author(s):  
M. Abdul-Niby ◽  
M. Alameen ◽  
O. Irscheid ◽  
M. Baidoun ◽  
H. Mourtada
Keyword(s):  
Obstacle Avoidance ◽  
Visually Impaired ◽  
Low Cost ◽  
Cost Effective ◽  
Obstacle Detection ◽  
Text To Speech ◽  
Voice Signal ◽  
Visually Impaired People ◽  
Impaired People ◽  
The Voice

In this paper, we present a low cost hands-free detection and avoidance system designed to provide mobility assistance for visually impaired people. An ultrasonic sensor is attached to the jacket of the user and detects the obstacles in front. The information obtained is transferred to the user through audio messages and also by a vibration. The range of the detection is user-defined. A text-to-speech module is employed for the voice signal. The proposed obstacle avoidance device is cost effective, easy to use and easily upgraded.

scholarly journals INFERRING THE SCALE AND CONTENT OF A MAP USING DEEP LEARNING

2020 ◽  
Vol XLIII-B4-2020 ◽  
pp. 17-24
Author(s):  
G. Touya ◽  
F. Brisebard ◽  
F. Quinton ◽  
A. Courtial
Keyword(s):  
Neural Networks ◽  
Deep Learning ◽  
Convolutional Neural Networks ◽  
Visually Impaired ◽  
The Other ◽  
Visually Impaired People ◽  
Impaired People ◽  
Tactile Maps

Abstract. Visually impaired people cannot use classical maps but can learn to use tactile relief maps. These tactile maps are crucial at school to learn geography and history as well as the other students. They are produced manually by professional transcriptors in a very long and costly process. A platform able to generate tactile maps from maps scanned from geography textbooks could be extremely useful to these transcriptors, to fasten their production. As a first step towards such a platform, this paper proposes a method to infer the scale and the content of the map from its image. We used convolutional neural networks trained with a few hundred maps from French geography textbooks, and the results show promising results to infer labels about the content of the map (e.g. ”there are roads, cities and administrative boundaries”), and to infer the extent of the map (e.g. a map of France or of Europe).

Sign in / Sign up

Export Citation Format

Share Document