Doing activities such as navigation is a big problem for people with visual impairment. It makes them inactive and isolates them from communicating with the people around them. A lot of technological interventions have been proposed to solve and overcome these problems. This paper proposes a solution to identify popular objects and avoid obstacles around them. YOLOv3 and Tiny-YOLO3 deep learning models are trained with multiple images containing obstacles that the visually impaired person will face indoors. The results show an average accuracy of 94.6% for object detection while using the YOLOv3 model, and 97.91% recognition accuracy is achieved for using the same model.