Multiple Objects Detection based on Improved Faster R-CNN
Qinghui Zhang, Chenxia Wan, Mingwei Jiang · 2017
Object detection is one of the hotspots in recent years. In order to solve those problems that many traditional methods exist such as single object detection and poor robustness detection, a multiple objects detection model based on the improved Faster R-CNN has been proposed. By introducing the inception network structure and changing the parameters such as anchor, the Faster R-CNN network model and the improved Faster R-CNN network model were structured and experimented on the PASCAL VOC2007 dataset, respectively. The experimental results showed that the improved Faster R-CNN network model have made up for the many shortcomings of the original model, especially for the images of the multiple objects. The improved network can not only fully extract the complex information in the images, but also identify the small objects that cannot be identified originally. And it is easy to distinguish the confused objects, so that the recognition rate of the final results has been improved to a certain degree than the previous.