Recognize Objects for Visually Impaired using Computer Vision
Deven Pawar1, Mihir Raul2, Pranav Raut3, Sharmila Gaikwad4

1Deven Pawar, U.G. Student Computer Department, Rajiv Gandhi Institute of Technology, Mumbai, India.
2Mihir Raul, U.G. Student Computer Department, Rajiv Gandhi Institute of Technology, Mumbai, India.
3Pranav Raut, U.G. Student Computer Department, Rajiv Gandhi Institute of Technology, Mumbai, India.
4Sharmila Gaikwad, Asst. Professor, Rajiv Gandhi Institute of Technology, Mumbai, India.
Manuscript received on February 28, 2020. | Revised Manuscript received on March 22, 2020. | Manuscript published on March 30, 2020. | PP: 5365-5369 | Volume-8 Issue-6, March 2020. | Retrieval Number: F9579038620/2020©BEIESP | DOI: 10.35940/ijrte.F9579.038620

Open Access | Ethics and Policies | Cite | Mendeley
© The Authors. Blue Eyes Intelligence Engineering and Sciences Publication (BEIESP). This is an open access article under the CC BY-NC-ND license (

Abstract: Visually impaired people are often unaware of dangers in front of them, even in familiar environments. Due to lack of vision either partial or complete, such people are highly dependent on the sense of hearing to perform day to day activities. One more form of vision impairment is colour blindness. Individuals with colour blindness find it hard to distinguish between colours. This research proposes making software that can help such individuals for solving the unawareness of the surrounding of the visually impaired people which allows them to have a greater awareness of their surroundings. The software needs an input device typically a camera and an audio feedback system. The camera will continuously capture images and the algorithm recognize the objects in the image and output the result using the audio feedback system. The system also proposes to include colour extraction to also correctly identify the colour of the object and a further addition is to identify individuals if enough datasets are provided. If any suspicious/dangerous objects detected in the surrounding the software will inform the user about the imminent danger. This study has analysed Faster R-CNN, SSD (Single Shot MultiBox Detector) and YOLO (You only look once) for their accuracy and rate of object detection. This research also studied different operating scenarios of the device which includes operation at night and operation in various orientations. The results of the object recognition system while using YOLO have an accuracy of 59.7% and 10fps during real-time operation, which is sufficient for assisting visually impaired people in realizing the types and localities of the objects around them.
Keywords: Computer Vision F-CNN, Raspberry-Pi, Visually Impaired, YOLO.
Scope of the Article: Data Management.