AI-based Object Detection for Assisting the Visually Impaired People

In recent decades, vision problems have become more common. One significant barrier is that people who are visually impaired generally require assistance performing their daily tasks. As a result, developing a technological solution to this challenge is critical. Implementing an automatic object identification system can enable visually impaired people to roam independently. The study employs two Artificial Intelligence (AI) models, a Region-based Fully Convolutional Network (RFCN) and a Mask Region-based Convolutional Neural Network (Mask RCNN), to detect objects and assess their performance. The evaluation demonstrates RFCN's outstanding performance with an impressive Mean Average Precision (mAP) score of 0.825. The Raspberry Pi processor is used as a vital component in the process of hardware development. The RFCN model is integrated into the processor to aid in the identification of objects collected by the system's pi-camera. The detected objects are communicated to users via auditory feedback. The hardware system also includes an ultrasonic sensor, which provides distance information for the detected objects. This full hardware solution is specifically developed to help visually impaired people overcome mobility issues in both indoor and outdoor settings. The suggested system seeks to improve the autonomy and safety of visually impaired individuals in their day-to-day activities by seamlessly mixing advanced technologies such as AI and embedded systems.