Ensemble of Deep Object Detectors for Page Object Detection

Document Imaging Understanding (DIU) is the process of converting all of the information content of a document image digital into an electronic format launched its reasonable content. We first evaluate different state-of-the-art object detection methods (Fast R-CNN and Faster R-CNN) for the task on POD dataset. We observe that each detection method is sensitive to certain objects. Therefore, we propose combining the detections of Fast RCNN and Faster RCNN in order to exploit the advantages of the two models. Through the extensive experiments, our proposed method significantly improves from 81.35% (Faster R-CNN VGG_CNN_M_1024) to 86.49% in terms of the mAP.