A Lightweight Region Proposal Network for Task Specific Applications

Quickly and cheaply finding areas of interest within an image can save computationally intensive image processing in a vision pipeline. Existing region proposal networks are either too general (finding all objects in an image) or too complex (providing fine-tuned bounding boxes for classification). We propose a straightforward region proposal network that simply scores parts of the image based on whether or not they contain an object of interest. This calculation can be carried out quickly and for many applications including autonomous driving only a small fraction of of the image area may contain objects of interest. We trained our network on an autonomous robot soccer dataset with similar characteristics to the popular KITTI autonomous driving dataset and achieved a recall greater than 95% while eliminating on average over 80% of the image area from further processing.

[1]  Ali Farhadi,et al.  YOLO9000: Better, Faster, Stronger , 2016, 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR).

[2]  C. Lawrence Zitnick,et al.  Edge Boxes: Locating Object Proposals from Edges , 2014, ECCV.

[3]  Bernt Schiele,et al.  What Makes for Effective Detection Proposals? , 2015, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[4]  David A. McAllester,et al.  Object Detection with Discriminatively Trained Part Based Models , 2010, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[5]  Paul A. Viola,et al.  Rapid object detection using a boosted cascade of simple features , 2001, Proceedings of the 2001 IEEE Computer Society Conference on Computer Vision and Pattern Recognition. CVPR 2001.

[6]  Koen E. A. van de Sande,et al.  Selective Search for Object Recognition , 2013, International Journal of Computer Vision.

[7]  Baolin Yin,et al.  Cracking BING and Beyond , 2014, BMVC.

[8]  Philip H. S. Torr,et al.  BING: Binarized normed gradients for objectness estimation at 300fps , 2019, Computational Visual Media.

[9]  Bernt Schiele,et al.  How good are detection proposals, really? , 2014, BMVC.

[10]  Andrew Zisserman,et al.  Very Deep Convolutional Networks for Large-Scale Image Recognition , 2014, ICLR.

[11]  Geoffrey E. Hinton,et al.  ImageNet classification with deep convolutional neural networks , 2012, Commun. ACM.

[12]  Andreas Geiger,et al.  Are we ready for autonomous driving? The KITTI vision benchmark suite , 2012, 2012 IEEE Conference on Computer Vision and Pattern Recognition.

[13]  Kaiming He,et al.  Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks , 2015, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[14]  Trevor Darrell,et al.  Rich Feature Hierarchies for Accurate Object Detection and Semantic Segmentation , 2013, 2014 IEEE Conference on Computer Vision and Pattern Recognition.

[15]  Simon O'Keeffe,et al.  A Benchmark Data Set and Evaluation of Deep Learning Architectures for Ball Detection in the RoboCup SPL , 2017, RoboCup.