The performance of Faster R-CNN algorithm on a dataset with poor capturing conditions

Abstract

In the past few years, computer vision algorithms have made a breakthrough in the field of object discovery, taking advantage of the significant development in computing capabilities and the availability of huge amounts of data with the emergence of many methods and techniques that have been used to achieve efficient results. In this research paper, we examine one of the most important object-detection algorithms, the Faster RCNN algorithm, and explore its ability and efficiency using low-quality or low-light image datasets captured in harsh conditions such as darkness or fog. Many research papers do not contain this kind of nature of images in their data, and their percentage is small in research, so this is a gap that we are trying to cover in this paper, at which we got a detection accuracy of mAP(50)=70.7%. The data contained ten capturing conditions (shadow, twilight, etc.) and 12 categories between human, animal, vehicle or furniture.