Executive Summary : | This project intends to improve the quality of life of visually impaired people by allowing them to recognize and identify a variety of real-world indoor items in aural mode, as well as the object's distance measurements, with minimal assistance. Object detection is a branch of computer vision (CV), which is an Artificial Intelligence field concerned with determining the existence, location, and identity of objects. The purpose of this project is to create a low-cost AI-based Smart Eye system that can identify and classify indoor items at a distance via aural mode for visually impaired persons. The Smart Eye's design incorporates computer vision, deep learning, and transfer learning. The system's four modules are object detection, categorization, distance measurement, and text-to-speech. Module one evaluates one-stage and two-stage object detection systems for theoretical and practical object recognition and localization research. By taking an input image and learning the class probabilities and bounding box coordinates, one–stage object detection approaches object recognition as a straightforward regression problem (SSD, YOLO). The two-stage object detection technique uses RCNN and FAST RCNN to locate plausible object regions. Deep learning and transfer learning are used in the second module to classify the discovered object. The third module comprises locating a spatial item and employing distance-measuring instruments to calculate the distance between the object and the individual. Finally, in the fourth module, the system uses text-to-speech conversion to offer audio output to the visually impaired person. This Smart AI Eye technology will boost the visually impaired person's ability to recognize items without aid by explaining the complete object existence in the focused zone as well as distinguishing specific things in aural mode. To give a low-cost deliverable, this smart Eye system will be developed as a mobile application. |