考虑数据分布损失的图像分割

doi:10.3778/j.issn.1002-8331.2307-0076

摘要/Abstract

摘要： 在图像分割任务中，损失函数的选择直接影响模型的收敛过程和最终精度。交叉熵损失函数（cross-entropy loss，CEL）具有稳定的收敛性，但在面临数据分布不均衡的情况下精度较低。Dice损失函数（Dice loss，DL）通过区域面积计算，在处理不均衡数据时能够获得较高的精度，但在多类别数据集上难以训练。为了解决这些问题，提出了一种的交叉熵损失函数（Dice cross-entropy loss，DCEL），它对正样本使用交叉熵计算损失值，而对负样本使用交叉熵与交并比（IOU）的乘积计算损失值。这样设计使得DCEL在多类别数据集上梯度与误差正相关，有利于模型的收敛性，并通过压缩负样本损失值提升对正样本的关注度，从而提升了图像分割算法的精度。DCEL的性能在ADE20k、PASCAL VOC、LoveDA和HRF数据集上进行了验证。

关键词: 损失函数, 图像分割, 不平衡数据, 深度学习

Abstract: The choice of loss function in image segmentation tasks directly impacts the convergence process and final accuracy of the models. Cross-entropy loss (CEL) exhibits stable convergence but achieves lower accuracy when dealing with imbalanced data distributions. Dice loss (DL), which calculates based on region overlap, achieves higher accuracy in handling imbalanced data but faces difficulties when applied to multi-class datasets. To address these issues, a modified loss function called Dice cross-entropy loss (DCEL) is proposed. DCEL computes the loss value using cross-entropy for positive samples and the product of cross-entropy and intersection over union (IOU) for negative samples. This design enables DCEL to have gradients positively correlated with errors on multi-class datasets, facilitating convergence. Furthermore, by compressing the loss value of negative samples, DCEL enhances the focus on positive samples, thereby improving the accuracy of image segmentation algorithms. The performance of DCEL is validated on ADE20k, PASCAL VOC, LoveDA, and HRF datasets.

Key words: loss function, image segmentation, imbalanced data, deep learning

张震, 彭景昊, 田鸿朋. 考虑数据分布损失的图像分割[J]. 计算机工程与应用, 2024, 60(19): 242-249.

ZHANG Zhen, PENG Jinghao, TIAN Hongpeng. Design of Loss Function Considering Data Distribution[J]. Computer Engineering and Applications, 2024, 60(19): 242-249.

参考文献

[1] MINAEE S, BOYKOV Y, PORIKLI F, et al. Image segmentation using deep learning: a survey[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2021, 44(7): 3523-3542.
[2] HU H, WEI F, HU H, et al. Semi-supervised semantic segmentation via adaptive equalization learning[C]//Advances in Neural Information Processing Systems, 2021, 34: 22106-22118.
[3] JADON S. A survey of loss functions for semantic segmentation[C]//Proceedings of the 2020 IEEE Conference on Computational Intelligence in Bioinformatics and Computational Biology (CIBCB), 2020: 1-7.
[4] PANG T, XU K, DONG Y, et al. Rethinking softmax cross-entropy loss for adversarial robustness[J]. arXiv:1905.10626, 2019.
[5] CUI Y, JIA M, LIN T Y, et al. Class-balanced loss based on effective number of samples[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2019: 9268-9277.
[6] MUKHOTI J, KULHARIA V, SANYAL A, et al. Calibrating deep neural networks using focal loss[C]//Advances in Neural Information Processing Systems, 2020, 33: 15288-15299.
[7] ZHAO R, QIAN B, ZHANG X, et al. Rethinking dice loss for medical image segmentation[C]//Proceedings of the 2020 IEEE International Conference on Data Mining (ICDM), 2020: 851-860.
[8] SALEHI S S M, ERDOGMUS D, GHOLIPOUR A. Tversky loss function for image segmentation using 3D fully convolutional deep networks[C]//Proceedings of the 8th International Workshop on Machine Learning in Medical Imaging, Quebec City, QC, Canada, 2017: 379-387.
[9] TAGHANAKI S A, ZHENG Y, ZHOU S K, et al. Combo loss: handling input and output imbalance in multi-organ segmentation[J]. Computerized Medical Imaging and Graphics, 2019, 75: 24-33.
[10] KAUR H, PANNU H S, MALHI A K. A systematic review on imbalanced data challenges in machine learning: applications and solutions[J]. ACM Computing Surveys (CSUR), 2019, 52(4): 1-36.
[11] YEUNG M, SALA E, SCH?NLIEB C B, et al. Unified Focal loss: generalising Dice and cross entropy-based losses to handle class imbalanced medical image segmentation[J]. Computerized Medical Imaging and Graphics, 2022, 95: 102026.
[12] ZHOU B, ZHAO H, PUIG X, et al. Semantic understanding of scenes through the ADE20k dataset[J]. International Journal of Computer Vision, 2019, 127: 302-321.
[13] SUDRE C H, LI W, VERCAUTEREN T, et al. Generalised dice overlap as a deep learning loss function for highly unbalanced segmentations[C]//Proceedings of the Third International Workshop on Deep Learning in Medical Image Analysis and Multimodal Learning for Clinical Decision Support, Québec City, QC, Canada, 2017: 240-248.
[14] SUTSKEVER I, MARTENS J, DAHL G, et al. On the importance of initialization and momentum in deep learning[C]//Proceedings of the International Conference on Machine Learning, 2013: 1139-1147.
[15] BHATTACHARYA P, ZHANG D, ZHANG L, et al. A new approach to automated retinal vessel segmentation using multiscale analysis[C]//Proceedings of the 18th International Conference on Pattern Recognition (ICPR’06), 2006: 77-80.
[16] WANG J, ZHENG Z, MA A, et al. LoveDA: a remote sensing land-cover dataset for domain adaptive semantic segmentation[J]. arXiv:2110.08733, 2021.
[17] TONG K, WU Y. Rethinking PASCAL-VOC and MS-COCO dataset for small object detection[J]. Journal of Visual Communication and Image Representation, 2023, 93: 103830.
[18] DAI Y, WU Y, ZHOU F, et al. Attentional local contrast networks for infrared small target detection[J]. IEEE Transactions on Geoscience and Remote Sensing, 2021, 59(11): 9813-9824.