Abstract
Noise robustness and hard example mining are two important aspects in object detection. A common view is that the two techniques are contradictory and they cannot be combined. In this paper, we show that there is a possibility to combine the best of two techniques. We find that, even using the hard example mining technique, recent deep neural network-based object detectors themselves have abilities to distinguish correct annotations and wrong annotations during the early stage of training. Based on this observation, we design a simple strategy to separate the wrong annotations from training data, reducing their loss weights and correcting their labels during training. The proposed method is simple, it doesn’t add any computational overhead during model inference. Moreover, the proposed method combines the hard example mining and noise resistance property in one model. Experiments on PASCAL VOC and DOTA datasets show that the proposed method not only archieves competitive performances on clean dataset, but also outperforms the baseline by a large margin when data contain severe noise.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Similar content being viewed by others
References
Arazo, E., Ortego, D., Albert, P., O’Connor, N., McGuinness, K.: Unsupervised label noise modeling and loss correction. In: ICML 2019: Thirty-Sixth International Conference on Machine Learning, pp. 312–321 (2019)
Chadwick, S., Newman, P.: Training object detectors with noisy data. In: 2019 IEEE Intelligent Vehicles Symposium (IV) (2019)
Cheng, G., Han, J., Zhou, P., Guo, L.: Multi-class geospatial object detection and geographic image classification based on collection of part detectors. Isprs J. Photogr. Remote Sens. 98(98), 119–132 (2014)
Dietterich, T.G., Lathrop, R.H., Lozano-Prez, T.: Solving the multiple instance problem with axis-parallel rectangles. Artif. Intell. 89(1), 31–71 (1997)
Dollár, P.: Focal loss for dense object detection. IEEE Trans. Pattern Anal. Mach. Intell. PP(99), 2999–3007 (2017)
Everingham, M., Gool, L., Williams, C.K., Winn, J., Zisserman, A.: The pascal visual object classes (VOC) challenge. Int. J. Comput. Vision 88(2), 303–338 (2010)
Gao, J., Wang, J., Dai, S., Li, L.J., Nevatia, R.: Note-RCNN: noise tolerant ensemble RCNN for semi-supervised object detection. In: 2019 IEEE/CVF International Conference on Computer Vision (ICCV), pp. 9507–9516 (2019)
Han, B., et al.: Co-teaching: robust training of deep neural networks with extremely noisy labels. arXiv preprint arXiv:1804.06872 (2018)
He, K., Zhang, X., Ren, S., Sun, J.: Deep residual learning for image recognition. In: 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pp. 770–778 (2016)
Hoffman, J., et al.: LSDA: large scale detection through adaptation. In: Advances in Neural Information Processing Systems, vol. 27, pp. 3536–3544 (2014)
Jiang, L., Zhou, Z., Leung, T., Li, L.J., Fei-Fei, L.: Learning data-driven curriculum for very deep neural networks on corrupted labels. In: ICML 2018: Thirty-fifth International Conference on Machine Learning (2018)
Lin, T.-Y., et al.: Microsoft COCO: common objects in context. In: Fleet, D., Pajdla, T., Schiele, B., Tuytelaars, T. (eds.) ECCV 2014. LNCS, vol. 8693, pp. 740–755. Springer, Cham (2014). https://doi.org/10.1007/978-3-319-10602-1_48
Shrivastava, A., Gupta, A., Girshick, R.: [IEEE 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) - Las Vegas, NV, USA (2016.6.27-2016.6.30)] 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) - training region-based object detectors with online hard example. In: IEEE Conference on Computer Vision & Pattern Recognition (2016)
Tang, Y., Wang, J., Gao, B., Dellandrea, E., Gaizauskas, R., Chen, L.: Large scale semi-supervised object detection using visual and semantic knowledge transfer. In: 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pp. 2119–2128 (2016)
Uijlings, J.R.R., Popov, S., Ferrari, V.: Revisiting knowledge transfer for training object class detectors. In: 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp. 1101–1110 (2018)
Wang, X., Shrivastava, A., Gupta, A.: A-fast-RCNN: hard positive generation via adversary for object detection. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp. 2606–2615 (2017)
Xia, G.S., et al.: DOTA: a large-scale dataset for object detection in aerial images. In: 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp. 3974–3983 (2018)
Zhang, X., Yang, Y., Feng, J.: Learning to localize objects with noisy labeled instances. AAAI 2019 : Thirty-Third AAAI Conference on Artificial Intelligence, vol. 33, no. 1, pp. 9219–9226 (2019)
Author information
Authors and Affiliations
Corresponding author
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2020 Springer Nature Switzerland AG
About this paper
Cite this paper
Hu, Z., Gao, K., Zhang, X., Dou, Z. (2020). Noise Resistant Focal Loss for Object Detection. In: Peng, Y., et al. Pattern Recognition and Computer Vision. PRCV 2020. Lecture Notes in Computer Science(), vol 12306. Springer, Cham. https://doi.org/10.1007/978-3-030-60639-8_10
Download citation
DOI: https://doi.org/10.1007/978-3-030-60639-8_10
Published:
Publisher Name: Springer, Cham
Print ISBN: 978-3-030-60638-1
Online ISBN: 978-3-030-60639-8
eBook Packages: Computer ScienceComputer Science (R0)