A data augmentation method based on cycle-consistent adversarial networks for fluorescence encoded microsphere image analysis
A data augmentation method based on cycle-consistent adversarial networks for fluorescence encoded microsphere image analysis
复制标题
一种基于循环一致对抗网络的荧光编码微球图像分析数据增强方法
DOI：
10.1016/j.sigpro.2019.02.028
复制
发表时间：
2019-08
影响因子：
4.4
通讯作者：
Tao Luo
中科院分区：
文献类型：
--
作者：
Zaifeng Shi;Minghe Liu;Qingjie Cao;Huizheng Ren;Tao Luo
监督学习的训练过程需要大量的训练数据才能达到令人满意的性能。然而，生物图像的获取和注释是昂贵的。提出了一种基于循环一致对抗网络（CycleGAN）的数据增强方法，并将其应用于荧光编码微球（FEM）图像分析的标注图像实例生成。在本文中，我们生成了大量的合成有限元图像和相应的注释，由计算机脚本。训练CycleGAN的前向生成器将合成图像转换到真实的图像域，用于Mask Region卷积神经网络（Mask R-CNN）的训练数据增强。我们评估了Mask R-CNN上不同大小的真实的/合成的/变换的FEM图像训练集的训练结果。在变换后的训练集上进行训练后，在0.50的间隔上的平均精度（AP.50）收敛到95.6%，而AP.75达到91.8%，比合成图像集高出约10%。实验结果证明了该方法在带注释生物图像增强中的有效性。
The training process of supervised learning requires a large amount of training data to achieve satisfactory performance. However, the acquisition and annotation of the biological images is costly. This paper presents a data augmentation method based on Cycle-Consistent Adversarial Networks (CycleGAN) and applied to the annotated image example generation of Fluorescence Encoded Microsphere (FEM) image analysis. In this paper, we generate a large number of synthetic FEM images and corresponding annotations by computer scripts. The forward generator from CycleGAN is trained to transform the synthetic images into the real image domain for the training data augmentation of the Mask Region Convolutional Neural Network (Mask R-CNN). We.evaluated the training results for different sizes of real/synthetic/transformed FEM image training sets on Mask R-CNN. After training on the transformed training set, the Average Precision at the Interval Over Union (IoU) of 0.50 (AP.50) converges to 95.6% and the AP.75 reaches 91.8%, which is about 10% higher than that of the synthetic image set. The experimental results demonstrate the effectiveness of this method in annotated biological image augmentation.
登录
查看更多内容
影响因子：
2.5
作者：
Schindelin J;Rueden CT;Hiner MC;Eliceiri KW
通讯作者：
Eliceiri KW
近红外遥控上转换纳米粒子：一种改进的活细胞染料标记工具
影响因子：
3.5
作者：
Wang, Huiquan;Li, Wei;Tan, Jian;Chang, Jin
通讯作者：
Chang, Jin
DOI：
10.1109/icassp.2014.6854671
发表时间：
2014-05
期刊：
2014 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)
影响因子：
--
作者：
Xiaodong Cui;Vaibhava Goel;Brian Kingsbury
通讯作者：
Xiaodong Cui;Vaibhava Goel;Brian Kingsbury
自动驾驶中目标检测的特定类别目标建议重新排序
影响因子：
6
作者：
Zhong Zhun;Lei Mingyi;Cao Donglin;Fan Jianping;Li Shaozi
通讯作者：
Li Shaozi
DOI：
10.1109/cvprw.2017.112
发表时间：
2017-07
期刊：
Conference on Computer Vision and Pattern Recognition Workshops. IEEE Computer Society Conference on Computer Vision and Pattern Recognition. Workshops
影响因子：
--
作者：
Hung J;Lopes SCP;Nery OA;Nosten F;Ferreira MU;Duraisingh MT;Marti M;Ravel D;Rangel G;Malleret B;Lacerda MVG;Rénia L;Costa FTM;Carpenter AE
通讯作者：
Carpenter AE