Comparing and contrasting the performance of single and two stage object detectors

Abstract

Object detectors are used in a wide array of different contexts and as such different implementations will vary in terms of performance and results. This study investigated how four object detectors perform on several datasets. More specifically and what is of particular interest would be the performance of these classifiers on remotely sensed images. The four object detectors that are ultimately implemented would be the Regions with Convolutional Neural Networks (RCNN), the Faster Regions with Convolutional Neural Networks (Faster RCNN), You Only Look Once (YOLO) and the Single Shot Multi-box Detector (SSD). In regards to remotely sensed images, it is ultimately found that the RCNN had the detection highest accuracy followed by Faster RCNN, YOLO and the SSD.