A High Level CNN-based App to Guide the People of Visually Diminished
Balamuralikrishna Thati, Lakshmi Devi Sunkavalli, Indraja Tanuri, Chaturya Sureddy, Venkata KrishnVamsi, Sowmya Koneru · 2023
For a human in order to lead a normal life five senses place a crucial role among them vision takes a most significant place. These days most people are facing problems to understand situations and difficulties around them due to blindness. They are not able to work independently. So, they want guidance from other people such as family members, friends etc. In addition, it will give the information about the direction of things and distance from the place where they are located. It notifies about the direction of the objects and the distance of the objects. So, our paper makes effort in enlarging the detection of objects through an application for blind people. To implement this essential component like camera, an audio device and an android application are required using all these components we developed an object detecting android application for detecting and identifying objects for blind people. To overcome this problem, we have proposed an android application to detect the object and provide an audio message. The main outcome of the paper is to detecting objects by using Android applications and to help vision challenging people in highly sophisticated and effective manner. This paper consists of different modules such as object detection and object localization along with distance calculation which is developed using OpenCV. This paper narrated using Convolution Neural Network Algorithm and it have given a tremendous result. The techniques adopted and implemented has leveraged an accuracy of 86.7%, whereas the data file contains more than 1100 brackets, a point-based denary approach using the supported features of service analysis is performed, The denary analysis represents the overall points of the systems in an surprising manner. It is observed that the system which is proposed has performed better than the existing systems that having total score of 9.6/10, where it is 8.5% more than the second-best.