Deep Learning based Object Detection Methods: A Review
Divya Mishra*
Published: March 29, 2022
DOI: 10.55162/MCET.02.027
Abstract
Recently, object detection has become one of the effective and popular trends in computer vision to deal with numerous applications such as in medical image processing of breast cancer, skin cancer, brain injuries, blood cells, and more. Also, it is used in video surveillance stations for real-time monitoring of crowd and anomaly detection. The application is widely used in satellite images and astronomy, fraud detection, and in the field of remote sensing to detect disaster-prone areas from satellite images so that important measures can be taken at the correct time to overcome or reduce the loss of life and property. For example, in medical applications earlier diagnosis procedures usually tend to figure out early diabetes, cancer and a few more diseases. Despite many existing object detection methods in state-of-the-art literature, it is getting harder to identify the best fit model for the specific application or dataset. Therefore, it is highly important and much needed to bring all the existing techniques to a single platform and mention their advantages and limitations. In this paper, a thorough literature review and comparison of various existing deep learning-based object detection methods are presented using three different parameters named mean average precision, frames per second, and data set used. Such information is useful for researchers and practitioners to identify the better approaches among the others easily according to the dataset in hand.