Refining faster-RCNN for accurate object detection

노명철(카카오), 이주영(카카오)




Object detector with region proposal networks such as Fast/Faster R-CNN have shown the state-ofthe art performance on several benchmarks. However, they have limited success for detecting small objects. We argue the limitation is related to insufficient performance of Fast R-CNN block in Faster R-CNN. In this paper, we propose a refining block for Fast R-CNN. We further merge the block and Faster R-CNN into a single network (RF-RCNN). The RF-RCNN was applied on plate and human detection in RoadView image that consists of high resolution street images (over 30M pixels). As a result, the RF-RCNN showed great improvement over the Faster-RCNN.

[ Figure 1 ] Illustrations of the Faster-RCNN and the proposed Refining Faster-RCNN