Detection and Content Retrieval of Object in an Image using YOLO

Abstract

It is easy for human beings to identify the object that is in an image. Even if the task is complex, human beings require only a minimal effort. Since computer vision is actually replicating human visual system, the same thing can be achieved in computers when they are trained with large amount of data, faster GPUs and many advanced algorithms. In general terms, Object detection can be defined as a technology that detects instances of object in images and videos by mimicking the human visual system functionalities. The motivation of the project is making the search process easier for the user i.e., if the object is very new for the user and he has no idea about it, he can upload a picture of that object and the algorithm will detect the object and gives a description about it. The objective of the project is to detect the object in an image, once the object is detected, the label i.e., the name of the detected object is searched in Wikipedia and few lines of description about that object is retrieved and printed. Also, the label is searched in google and the URL of the top pages with content related to the label are also displayed. The detection of object in an image is done using YOLO (You Only Look Once) algorithm with pre-trained weights. Previous methods for object detection, like R-CNN and its variations, used a pipeline to perform this task in multiple steps. This can take some time for execution, complex optimization may be involved because individual training of components is required. YOLO, does it all fastly with a single neural network. Hence, YOLO is preferred.