Yuxin Wu

Kaiming He


Batch Normalization (BN) is a milestone technique in the development of deep learning, enabling various networks to train. However, normalizing along the batch dimension introduces problems—BN’s error increases rapidly when the batch size becomes smaller, caused by inaccurate batch statistics estimation. This limits BN’s usage for training larger models and transferring features to computer vision tasks including detection, segmentation, and video, which require small batches constrained by memory consumption. In this paper, we present Group Normalization (GN) as a simple alternative to BN. GN divides the channels into groups and computes within each group the mean and variance for normalization. GN’s computation is independent of batch sizes, and its accuracy is stable in a wide range of batch sizes. On ResNet-50 trained in ImageNet, GN has 10.6% lower error than its BN counterpart when using a batch size of 2; when using typical batch sizes, GN is comparably good with BN and outperforms other normalization variants. Moreover, GN can be naturally transferred from pre-training to fine-tuning. GN can outperform its BN-based counterparts for object detection and segmentation in COCO, and for video classification in Kinetics, showing that GN can effectively replace the powerful BN in a variety of tasks. GN can be easily implemented by a few lines of code.



1. Ioffe, S., Szegedy, C.: Batch normalization: accelerating deep network training by reducing internal covariate shift. In: ICML (2015)Google Scholar
2. Szegedy, C., Vanhoucke, V., Ioffe, S., Shlens, J., Wojna, Z.: Rethinking the inception architecture for computer vision. In: CVPR (2016)Google Scholar
3. He, K., Zhang, X., Ren, S., Sun, J.: Deep residual learning for image recognition. In: CVPR (2016)Google Scholar
4. Silver, D., et al.: Mastering the game of go without human knowledge. Nature 550(7676), 354–359 (2017)CrossRefGoogle Scholar
5. Szegedy, C., Ioffe, S., Vanhoucke, V.: Inception-v4, inception-resnet and the impact of residual connections on learning. In: ICLR Workshop (2016)Google Scholar
6. Huang, G., Liu, Z., van der Maaten, L., Weinberger, K.Q.: Densely connected convolutional networks. In: CVPR (2017)Google Scholar
7. Xie, S., Girshick, R., Dollár, P., Tu, Z., He, K.: Aggregated residual transformations for deep neural networks. In: CVPR (2017)Google Scholar
8. Girshick, R.: Fast R-CNN. In: ICCV (2015)Google Scholar
9. Ren, S., He, K., Girshick, R., Sun, J.: Faster R-CNN: towards real-time object detection with region proposal networks. In: NIPS (2015)Google Scholar
10. He, K., Gkioxari, G., Dollár, P., Girshick, R.: Mask R-CNN. In: ICCV (2017)Google Scholar
11. Long, J., Shelhamer, E., Darrell, T.: Fully convolutional networks for semantic segmentation. In: CVPR (2015)Google Scholar
12. Tran, D., Bourdev, L., Fergus, R., Torresani, L., Paluri, M.: Learning spatiotemporal features with 3D convolutional networks. In: ICCV (2015)Google Scholar
13. Carreira, J., Zisserman, A.: Quo vadis, action recognition? A new model and the kinetics dataset. In: CVPR (2017)Google Scholar
14. Lowe, D.G.: Distinctive image features from scale-invariant keypoints. In: IJCV (2004)MathSciNetCrossRefGoogle Scholar
15. Dalal, N., Triggs, B.: Histograms of oriented gradients for human detection. In: CVPR (2005)Google Scholar
16. Russakovsky, O., et al.: ImageNet large scale visual recognition challenge. In: IJCV (2015)Google Scholar
17. Ba, J.L., Kiros, J.R., Hinton, G.E.: Layer normalization (2016). arXiv:1607.06450
18. Ulyanov, D., Vedaldi, A., Lempitsky, V.: Instance normalization: the missing ingredient for fast stylization (2016). arXiv:1607.08022
19. Salimans, T., Kingma, D.P.: Weight normalization: a simple reparameterization to accelerate training of deep neural networks. In: NIPS (2016)Google Scholar
20. Lin, T.-Y., et al.: Microsoft COCO: common objects in context. In: Fleet, D., Pajdla, T., Schiele, B., Tuytelaars, T. (eds.) ECCV 2014, Part V. LNCS, vol. 8693, pp. 740–755. Springer, Cham (2014). Scholar
21. Kay, W., et al.: The kinetics human action video dataset (2017). arXiv:1705.06950
22. Rumelhart, D.E., Hinton, G.E., Williams, R.J.: Learning representations by back-propagating errors. Nature 323, 533–536 (1986)CrossRefGoogle Scholar
23. Hochreiter, S., Schmidhuber, J.: Long short-term memory. Neural Comput. 9(8), 1735–1780 (1997)CrossRefGoogle Scholar
24. Goodfellow, I., et al.: Generative adversarial nets. In: NIPS (2014)Google Scholar
25. Isola, P., Zhu, J.Y., Zhou, T., Efros, A.A.: Image-to-image translation with conditional adversarial networks. In: CVPR (2017)Google Scholar
26. Lyu, S., Simoncelli, E.P.: Nonlinear image representation using divisive normalization. In: CVPR (2008)Google Scholar
27. Jarrett, K., Kavukcuoglu, K., LeCun, Y., et al.: What is the best multi-stage architecture for object recognition? In: ICCV (2009)Google Scholar
28. Krizhevsky, A., Sutskever, I., Hinton, G.: Imagenet classification with deep convolutional neural networks. In: NIPS (2012)Google Scholar
29. Zeiler, M.D., Fergus, R.: Visualizing and understanding convolutional networks. In: Fleet, D., Pajdla, T., Schiele, B., Tuytelaars, T. (eds.) ECCV 2014, Part I. LNCS, vol. 8689, pp. 818–833. Springer, Cham (2014). Scholar
30. Sermanet, P., Eigen, D., Zhang, X., Mathieu, M., Fergus, R., LeCun, Y.: Overfeat: integrated recognition, localization and detection using convolutional networks. In: ICLR (2014)Google Scholar
31. Szegedy, C., et al.: Going deeper with convolutions. In: CVPR (2015)Google Scholar
32. Rebuffi, S.A., Bilen, H., Vedaldi, A.: Learning multiple visual domains with residual adapters. In: NIPS (2017)Google Scholar
33. Arpit, D., Zhou, Y., Kota, B., Govindaraju, V.: Normalization propagation: a parametric technique for removing internal covariate shift in deep networks. In: ICML (2016)Google Scholar
34. Ren, M., Liao, R., Urtasun, R., Sinz, F.H., Zemel, R.S.: Normalizing the normalizers: comparing and extending network normalization schemes. In: ICLR (2017)Google Scholar
35. Ioffe, S.: Batch renormalization: towards reducing minibatch dependence in batch-normalized models. In: NIPS (2017)Google Scholar
36. Peng, C., et al.: MegDet: a large mini-batch object detector. In: CVPR (2018)Google Scholar
37. Dean, J., et al.: Large scale distributed deep networks. In: NIPS (2012)Google Scholar
38. Howard, A.G., et al.: MobileNets: efficient convolutional neural networks for mobile vision applications (2017). arXiv:1704.04861
39. Chollet, F.: Xception: deep learning with depthwise separable convolutions. In: CVPR (2017)Google Scholar
40. Zhang, X., Zhou, X., Lin, M., Sun, J.: ShuffleNet: an extremely efficient convolutional neural network for mobile devices. In: CVPR (2018)Google Scholar
41. Oliva, A., Torralba, A.: Modeling the shape of the scene: a holistic representation of the spatial envelope. In: IJCV (2001)Google Scholar
42. Jegou, H., Douze, M., Schmid, C., Perez, P.: Aggregating local descriptors into a compact image representation. In: CVPR (2010)Google Scholar
43. Perronnin, F., Dance, C.: Fisher kernels on visual vocabularies for image categorization. In: CVPR (2007)Google Scholar
44. Dieleman, S., De Fauw, J., Kavukcuoglu, K.: Exploiting cyclic symmetry in convolutional neural networks. In: ICML (2016)Google Scholar
45. Cohen, T., Welling, M.: Group equivariant convolutional networks. In: ICML (2016)Google Scholar
46. Heeger, D.J.: Normalization of cell responses in cat striate cortex. Vis. Neurosci. 9(2), 181–197 (1992)MathSciNetCrossRefGoogle Scholar
47. Schwartz, O., Simoncelli, E.P.: Natural signal statistics and sensory gain control. Nat. Neurosci. 4(8), 819 (2001)CrossRefGoogle Scholar
48. Simoncelli, E.P., Olshausen, B.A.: Natural image statistics and neural representation. Ann. Rev. Neurosci. 24(1), 1193–1216 (2001)CrossRefGoogle Scholar
49. Carandini, M., Heeger, D.J.: Normalization as a canonical neural computation. Nat. Rev. Neurosci. 13(1), 51 (2012)CrossRefGoogle Scholar
50. Paszke, A., et al.: Automatic differentiation in pytorch (2017)Google Scholar
51. Abadi, M., et al.: Tensorflow: a system for large-scale machine learning. In: Operating Systems Design and Implementation (OSDI) (2016)Google Scholar
52. Gross, S., Wilber, M.: Training and investigating Residual Nets (2016).
53. He, K., Zhang, X., Ren, S., Sun, J.: Delving deep into rectifiers: surpassing human-level performance on imagenet classification. In: ICCV (2015)Google Scholar
54. Goyal, P., et al.: Accurate, large minibatch SGD: training ImageNet in 1 hour (2017). arXiv:1706.02677
55. Krizhevsky, A.: One weird trick for parallelizing convolutional neural networks (2014). arXiv:1404.5997
56. Bottou, L., Curtis, F.E., Nocedal, J.: Optimization methods for large-scale machine learning (2016). arXiv:1606.04838
57. Simonyan, K., Zisserman, A.: Very deep convolutional networks for large-scale image recognition. In: ICLR (2015)Google Scholar
58. Lin, T.Y., Goyal, P., Girshick, R., He, K., Dollár, P.: Focal loss for dense object detection. In: ICCV (2017)Google Scholar
59. Girshick, R., Radosavovic, I., Gkioxari, G., Dollár, P., He, K.: Detectron (2018).
60. Lin, T.Y., Dollár, P., Girshick, R., He, K., Hariharan, B., Belongie, S.: Feature pyramid networks for object detection. In: CVPR (2017)Google Scholar
61. Ren, S., He, K., Girshick, R., Zhang, X., Sun, J.: Object detection networks on convolutional feature maps. TPAMI 39(7), 1476–1481 (2017)CrossRefGoogle Scholar
62. Li, Z., Peng, C., Yu, G., Zhang, X., Deng, Y., Sun, J.: DetNet: a backbone network for object detection (2018). arXiv:1804.06215
63. Wang, X., Girshick, R., Gupta, A., He, K.: Non-local neural networks. In: CVPR (2018)Google Scholar

The SELF Institute