S-OHEM: Stratified Online Hard Example Mining for Object Detection
作者： Zhaoning Zhang , Dongsheng Li , Minne Li , Xinyuan Chen , Hao Yu
DOI:
关键词:
摘要: One of the major challenges in object detection is to propose detectors with highly accurate localization objects. The online sampling high-loss region proposals (hard examples) uses multitask loss equal weight settings across all types (e.g, classification and localization, rigid non-rigid categories) ignores influence different distributions throughout training process, which we find essential efficacy. In this paper, present Stratified Online Hard Example Mining (S-OHEM) algorithm for higher efficiency accuracy detectors. S-OHEM exploits OHEM stratified sampling, a widely-adopted technique, choose examples according during hard example mining, thus enhance performance We show through systematic experiments that yields an average precision (AP) improvement 0.5% on categories PASCAL VOC 2007 both IoU threshold 0.6 0.7. For KITTI 2012, results same metric are 1.6%. Regarding mean (mAP), relative increase 0.3% (1% 0.5%) observed VOC07 (KITTI12) using set threshold. Also, easy integrate existing region-based capable acting post-recognition level regressors.
arxiv.org
core.ac.uk参考文章(26)
Diederik P. Kingma, Jimmy Ba, Adam: A Method for Stochastic Optimization arXiv: Learning. ,(2014)
Tomaso A. Poggio, Kah Kay Sung, Learning and example selection for object and pattern detection Massachusetts Institute of Technology. ,(1996)
Kaiming He, Xiangyu Zhang, Shaoqing Ren, Jian Sun, Delving Deep into Rectifiers: Surpassing Human-Level Performance on ImageNet Classification international conference on computer vision. pp. 1026- 1034 ,(2015) , 10.1109/ICCV.2015.123
Karen Simonyan, Andrew Zisserman, Very Deep Convolutional Networks for Large-Scale Image Recognition computer vision and pattern recognition. ,(2014)
Jonathan Long, Evan Shelhamer, Trevor Darrell, Fully convolutional networks for semantic segmentation computer vision and pattern recognition. pp. 3431- 3440 ,(2015) , 10.1109/CVPR.2015.7298965
Spyros Gidaris, Nikos Komodakis, Object Detection via a Multi-region and Semantic Segmentation-Aware CNN Model international conference on computer vision. pp. 1134- 1142 ,(2015) , 10.1109/ICCV.2015.135
Yu Xiang, Roozbeh Mottaghi, Silvio Savarese, Beyond PASCAL: A benchmark for 3D object detection in the wild workshop on applications of computer vision. pp. 75- 82 ,(2014) , 10.1109/WACV.2014.6836101
Ilya Sutskever, Geoffrey Hinton, Alex Krizhevsky, Ruslan Salakhutdinov, Nitish Srivastava, Dropout: a simple way to prevent neural networks from overfitting Journal of Machine Learning Research. ,vol. 15, pp. 1929- 1958 ,(2014)
M Sai Praneeth, Xudong Peng, Alice Li, Shahrzad Hosseini Vajargah, Going deeper with convolutions computer vision and pattern recognition. pp. 1- 9 ,(2015) , 10.1109/CVPR.2015.7298594
Ross Girshick, Jeff Donahue, Trevor Darrell, Jitendra Malik, Rich Feature Hierarchies for Accurate Object Detection and Semantic Segmentation computer vision and pattern recognition. pp. 580- 587 ,(2014) , 10.1109/CVPR.2014.81