Real Time Object Detection Using Faster R-CNNs and RPNs

Abstract

In this project we investigate serveral methods for Real-time object detection. The first one is the method proposed by Ren [1]. They propose the idea of Region Proposal Networks (RPNs) that share conv layers with state-of-the-art object detection networks such as Fast R-CNNs. Sharing the parameters reduces the proposal time for the images.

References

1. Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks, [paper] [code].