Object Detection Technique Using a Fast R-CNN Based Technique
Amlan Dutta · International Journal for Research in Applied Science and Engineering Technology · 2022
Abstract: Object detection stands as the current interest in locating the objects. Reducing the run time of the algorithms holds a challenge in this field. Deep Convolutional Neural Network (CNN) like SPPnet, Fast R-CNN etc. had been effectively proposed to reduce the execution time. Encouraged by these networks, a Region Program Network (RPN) has been proposed here to portray features of images with less computational cost and time. This network is capable of predicting the bounds of an object and generates scores at each position. In the proposed network, Fast R-CNN is also integrated with RPN so that exceptional region programs can be generated, and entire network can easily understand which area should be focused. The proposed network has been tested on MS COCO, PASCAL VOC 2007, 2012 datasets. Results infer that the proposed RPN along with Fast R-CNN outperforms some popular models like deep VGG-16 [3], ImageNet, YOLO etc. in terms of all the accuracy parameters.