Özet:

Abstract With the increasing automation in today’s world, the need for finding and labelling objects in images and videos has grown exponentially. Be it managing traffic, self-driving cars or medical imaging, object detection is being used everywhere around us. Traditional methods for object detection, like SIFT or HOG features, are efficient but no longer compatible for today's needs as the processing of images needed are in real time that can not be done by these methods. These methods also make the procedure of training and preparing our model really complex and can only work with well-lit, front-faced, full-picture images of objects which is not always possible to achieve. So, the deep learning methods for object detection, like R-CNN, YOLO or RetinaNet, were introduced.These methods are being used worldwide to detect objects and make object detection automated and simpler. In this paper, we provide a review on both machine learning and deep learning approaches for object detection. Our review begins with an introduction to object detection, then we focus on all the methods used for object detection - machine learning approach and deep learning approach. Then we move on to all the advantages, challenges and applications of object detection. To conclude it, we mentioned the future scopes everyone can look forward to.

Anahtar Kelimeler: