Suppr超能文献

基于阈值优化卷积神经网络的自动图像标注方法。

Automatic image annotation method based on a convolutional neural network with threshold optimization.

机构信息

School of Computer Science & Technology, Taiyuan University of Science and Technology, Taiyuan, China.

Department of Computer Science & Technology, Xinzhou Teachers University, Xinzhou, China.

出版信息

PLoS One. 2020 Sep 23;15(9):e0238956. doi: 10.1371/journal.pone.0238956. eCollection 2020.

Abstract

In this study, a convolutional neural network with threshold optimization (CNN-THOP) is proposed to solve the issue of overlabeling or downlabeling arising during the multilabel image annotation process in the use of a ranking function for label annotation along with prediction probability. This model fuses the threshold optimization algorithm to the CNN structure. First, an optimal model trained by the CNN is used to predict the test set images, and batch normalization (BN) is added to the CNN structure to effectively accelerate the convergence speed and obtain a group of prediction probabilities. Second, threshold optimization is performed on the obtained prediction probability to derive an optimal threshold for each class of labels to form a group of optimal thresholds. When the prediction probability for this class of labels is greater than or equal to the corresponding optimal threshold, this class of labels is used as the annotation result for the image. During the annotation process, the multilabel annotation for the image to be annotated is realized by loading the optimal model and the optimal threshold. Verification experiments are performed on the MIML, COREL5K, and MSRC datasets. Compared with the MBRM, the CNN-THOP increases the average precision on MIML, COREL5K, and MSRC by 27%, 28% and 33%, respectively. Compared with the E2E-DCNN, the CNN-THOP increases the average recall rate by 3% on both COREL5K and MSRC. The most precise annotation effect for CNN-THOP is observed on the MIML dataset, with a complete matching degree reaching 64.8%.

摘要

在这项研究中,提出了一种具有阈值优化的卷积神经网络(CNN-THOP),以解决在使用排序函数进行标签注释以及预测概率的多标签图像注释过程中出现的过度标记或欠标记问题。该模型将阈值优化算法融合到 CNN 结构中。首先,使用经过 CNN 训练的最优模型对测试集图像进行预测,并在 CNN 结构中添加批量归一化(BN),以有效加快收敛速度并获得一组预测概率。其次,对获得的预测概率进行阈值优化,为每个标签类导出最优阈值,形成一组最优阈值。当该类标签的预测概率大于或等于相应的最优阈值时,将该类标签用作图像的注释结果。在注释过程中,通过加载最优模型和最优阈值来实现要注释的图像的多标签注释。在 MIML、COREL5K 和 MSRC 数据集上进行验证实验。与 MBRM 相比,CNN-THOP 分别将 MIML、COREL5K 和 MSRC 的平均精度提高了 27%、28%和 33%。与 E2E-DCNN 相比,CNN-THOP 在 COREL5K 和 MSRC 上的平均召回率分别提高了 3%。CNN-THOP 在 MIML 数据集上观察到最精确的注释效果,完全匹配度达到 64.8%。

https://cdn.ncbi.nlm.nih.gov/pmc/blobs/0004/7511011/58dc41551602/pone.0238956.g001.jpg

相似文献

1
Automatic image annotation method based on a convolutional neural network with threshold optimization.
PLoS One. 2020 Sep 23;15(9):e0238956. doi: 10.1371/journal.pone.0238956. eCollection 2020.
2
A localization strategy combined with transfer learning for image annotation.
PLoS One. 2021 Dec 8;16(12):e0260758. doi: 10.1371/journal.pone.0260758. eCollection 2021.
3
CM-supplement network model for reducing the memory consumption during multilabel image annotation.
PLoS One. 2020 Jun 1;15(6):e0234014. doi: 10.1371/journal.pone.0234014. eCollection 2020.
4
Application of high resolution computed tomography image assisted classification model of middle ear diseases based on 3D-convolutional neural network.
Zhong Nan Da Xue Xue Bao Yi Xue Ban. 2022 Aug 28;47(8):1037-1048. doi: 10.11817/j.issn.1672-7347.2022.210704.
5
MAIA-A machine learning assisted image annotation method for environmental monitoring and exploration.
PLoS One. 2018 Nov 16;13(11):e0207498. doi: 10.1371/journal.pone.0207498. eCollection 2018.
6
An Improved Convolutional Neural Network Algorithm and Its Application in Multilabel Image Labeling.
Comput Intell Neurosci. 2019 Jul 4;2019:2060796. doi: 10.1155/2019/2060796. eCollection 2019.
8
A CAD system for pulmonary nodule prediction based on deep three-dimensional convolutional neural networks and ensemble learning.
PLoS One. 2019 Jul 12;14(7):e0219369. doi: 10.1371/journal.pone.0219369. eCollection 2019.
10
FlyIT: Drosophila Embryogenesis Image Annotation based on Image Tiling and Convolutional Neural Networks.
IEEE/ACM Trans Comput Biol Bioinform. 2021 Jan-Feb;18(1):194-204. doi: 10.1109/TCBB.2019.2935723. Epub 2021 Feb 3.

引用本文的文献

1
Advances and challenges in automated malaria diagnosis using digital microscopy imaging with artificial intelligence tools: A review.
Front Microbiol. 2022 Nov 15;13:1006659. doi: 10.3389/fmicb.2022.1006659. eCollection 2022.

本文引用的文献

1
Receptive fields, binocular interaction and functional architecture in the cat's visual cortex.
J Physiol. 1962 Jan;160(1):106-54. doi: 10.1113/jphysiol.1962.sp006837.

文献AI研究员

20分钟写一篇综述,助力文献阅读效率提升50倍。

立即体验

用中文搜PubMed

大模型驱动的PubMed中文搜索引擎

马上搜索

文档翻译

学术文献翻译模型,支持多种主流文档格式。

立即体验