Online Hard Example Mining for Training One-Stage Object Detectors

KIPS Transactions on Software and Data Engineering, Vol. 7, No.5, pp.195-204, May 2018
10.3745/KTSDE.2018.7.5.195, Full Text

Abstract

In this paper, we propose both a new loss function and an online hard example mining scheme for improving the performance of single-stage object detectors which use deep convolutional neural networks. The proposed loss function and the online hard example mining scheme can not only overcome the problem of imbalance between the number of annotated objects and the number of background examples, but also improve the localization accuracy of each object. Therefore, the loss function and the mining scheme can provide intrinsically fast single-stage detectors with detection performance higher than or similar to that of two-stage detectors. In experiments conducted with the PASCAL VOC 2007 benchmark dataset, we show that the proposed loss function and the online hard example mining scheme can improve the performance of single-stage object detectors.


Statistics

Show / Hide Statistics

Statistics (Cumulative Counts from October 15, 2016)

Multiple requests among the same browser session are counted as one view. If you mouse over a chart, the values of data points will be shown.


Cite this paper

[KIPS Transactions Style]
I. Kim, "Online Hard Example Mining for Training One-Stage Object Detectors," KIPS Transactions on Software and Data Engineering, Vol.7, No.5, pp.195-204, 2018, DOI: 10.3745/KTSDE.2018.7.5.195.

[IEEE Style]
Incheol Kim, "Online Hard Example Mining for Training One-Stage Object Detectors," KIPS Transactions on Software and Data Engineering, vol. 7, no. 5, pp. 195-204, 2018. DOI: 10.3745/KTSDE.2018.7.5.195.

[ACM Style]
Kim, I. 2018. Online Hard Example Mining for Training One-Stage Object Detectors. KIPS Transactions on Software and Data Engineering, 7, 5, (2018), 195-204. DOI: 10.3745/KTSDE.2018.7.5.195.