Evaluating the State-of-the-Art Deep Learning Models for Object Recognition: Focusing on Features, Deep Models and Backbones

Abstract

Artificial intelligence (AI) is now the most popular method for gathering insights about the actual world from a wide variety of data sources. Finding the pattern in the studied data is the primary objective. The statistical methods or specialized filters are used in the next phase, which is extracting representative characteristics. Recently, advancements in deep learning models have allowed computers to recognize and locate objects in photos and videos with unprecedented precision and speed. In this research, we compare and contrast many state-of-the-art deep learning models in both methods, including Fast RCNN, Faster RCNN, RetinaNet, and YOLO. A few computer vision tasks are also covered, with a study of the underlying structures for each discussed.