Improved Semantic Segmentation Model and Its Application

doi:10.3778/j.issn.1002-8331.2210-0032

Abstract

Abstract: The training of semantic segmentation model requires complicated manual labeling, and there are also some problems in the construction and operation of semantic segmentation model, such as determining its hyperparameters and becoming bloated. To solve these problems, this paper proposes a label generation method based on heat map generated by ground truth box, which simplifies the manual labeling process of semantic segmentation training labels. A neural architecture search method with lower hardware requirements is proposed, which based on the differentiable neural architecture search method. By this method, the improved semantic segmentation model which contains a new feature pyramid is constructed. Tested on the helmet and mask detection datasets, compared with U-NET, FPN and other models, the new model takes the advantages in the number of parameters, calculation speed and accuracy.

Key words: semantic segmentation model, neural architecture search, feature pyramid structure, helmets and masks detection

摘要： 训练语义分割网络模型需要较为繁琐的人工标注作为训练标签，同时语义分割模型在构建和运行过程中也存在超参数较难确定以及模型过于庞大等问题。为解决这类问题，提出了一种基于标注框生成热点图的标签生成方法，简化了语义分割训练标签的人工标注过程。以及在可微分神经网络结构搜索方法的基础上提出了一种对硬件要求更低的神经网络结构搜索方法，并基于此种方法改进了特征金字塔结构，构建了一个改进的语义分割模型，并在安全帽与口罩检测数据集上进行了试验。与U-Net、FPN等模型比较，新的模型在参数量、计算速度以及精确度上都更有优势。

关键词: 语义分割模型, 神经网络结构搜索, 特征金字塔结构, 安全帽与口罩检测

WANG Yaowen, CHENG Junsheng, YANG Yu. Improved Semantic Segmentation Model and Its Application[J]. Computer Engineering and Applications, 2024, 60(2): 337-343.

王耀文, 程军圣, 杨宇. 改进的语义分割模型及其应用[J]. 计算机工程与应用, 2024, 60(2): 337-343.

References

[1] REDMON J, DIVVALA S, GIRSHICK R, et al. You only look once: unified, real-time object detection[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2016: 779-788.
[2] REDMON J, FARHADI A. YOLO9000: better, faster, stronger[C]//IEEE Conference on Computer Vision & Pattern Recognition, 2017: 6517-6525.
[3] REDMON J, FARHADI A. YOLOv3: an incremental improvement[C]//2018 IEEE Conference on Computer Vision and Pattern Recognition, 2018.
[4] LIU W, ANGUELOV D, ERHAN D, et al. SSD: single shot multibox detector[C]//European Conference on Computer Vision. Cham: Springer, 2016: 21-37.
[5] REN S, HE K, GIRSHIVK R, et al. Faster R-CNN: towards real-time object detection with region proposal networks[J]. In IEEE Transactions on Pattern Analysis and Machine Intelligence, 2017, 39(6): 1137-1149.
[6] LONG J, SHELHAMER E, DARRELL T. Fully convolutional networks forsemantic segmentation[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2015: 3431-3440.
[7] RONNEBERGER O, FISCHER P, BROX T. U-NET: convolutional networks for biomedical image segmentation[C]//International Conference on Medical Image Computingand Computer-Assisted Intervention. Cham: Springer, 2015: 234-241.
[8] LIN T Y, DOLLAR P, GIRSHICK R, et al. Feature pyramid networks for object detection[J]. 2017 IEEE Conference on Computer Vision and Pattern Recognition(CVPR), Honolulu, 2017: 936-944.
[9] HOWARD A G, ZHU M, CHEN B, et al. Mobilenets: efficient convolutional neural networks for mobile vision applications[C]//IEEE Conference on Computer Vision and Pattern Recognition, Hawaii, 2017: 1-9.
[10] SANDLER M, HOWARD A, ZHU M, et al. MobileNet V2: inverted residuals and linear bottlenecks[C]//IEEE Conference on Computer Vision and Pattern Recognition, Salt Lake City, 2018: 4510-4520.
[11] HOWARD A, SANDLER M, CHU G, et al. Searching for MobileNetV3[C]//Proceedings of the IEEE International Conference on Computer Vision, 2019.
[12] BARRET Z, QUOC V L. Neural architecture search with reinforcement learning[J]. arXiv:1611.01578, 2016.
[13] WU B C, DAI X L, ZHANG P Z, et al. FBNet: hardware-aware efficient ConvNet design via differentiable neural architecture search[C]//Proceddings of the CVPR, 2019.
[14] 张立艺, 武文红, 牛恒茂, 等. 深度学习中的安全帽检测算法应用研究综述[J]. 计算机工程与应用, 2022, 58(16): 1-17.
ZHANG L Y, WU W H, NIU H M, et al. Summary of application research on helmet detection algorithm based on deep learning[J]. Computer Engineering and Applications, 2022, 58(16): 1-17.
[15] ZHOU X, WANG D, KRHENBUHL P. Objects as points[J]. arXiv:1904.07850,2019.
[16] HU J, SHEN L, SUN G. Squeeze-and-excitation networks[J]. arXiv:1709.01507,2017.