U-Net-Based Segmentation of Microscopic Images of Colorants and Simplification of Labeling in the Learning Process

Abstract

Colored product textures correspond to particle size distributions. The microscopic images of colorants must be divided into regions to determine the particle size distribution. The conventional method used for this process involves manually dividing images into areas, which may be inefficient. In this paper, we have overcome this issue by developing two different modified architectures of U-Net convolution neural networks to automatically determine the particle sizes. To develop these modified architectures, a significant amount of ground truth data must be prepared to train the U-Net, which is difficult for big data as the labeling is performed manually. Therefore, we also aim to reduce this process by using incomplete labeling data. The first objective of this study is to determine the accuracy of our modified U-Net architectures for this type of image. The second objective is to reduce the difficulty of preparing the ground truth data by testing the accuracy of training on incomplete labeling data. The results indicate that efficient segmentation can be realized using our modified U-Net architectures, and the generation of ground truth data can be simplified. This paper presents a preliminary study to improve the efficiency of determining particle size distributions with incomplete labeling data.

Keywords: U-Net; colorants; deep learning; labeling; particle size distribution; segmentation.

Figure 1

Sample dataset. Input data for a (a) small and (b) large number of particles. Output data for a (c) small and (d) large number of particles.

Figure 2

Conventional U-Net.

Figure 3

Improved U-Net (U-Net #1) with fewer channels than the conventional architecture.

Figure 4

Feature maps of the deep layer (a) before (conventional U-Net) and (b) after (U-Net #1) the number of channels was decreased.

Figure 5

Part of feature maps of the hidden layer in U-Net #1.

Figure 6

Improved U-Net (U-Net #2): Deleted two skip connections.

Figure 7

Sample incomplete labeling images with (a) small number of particles; (b) large number of particles.

Figure 8

Results of image processing after segmentation (small number of particles). (a) Input data; (b) Output data (ground truth). Results obtained using (c) conventional U-Net; (d) U-Net #1; (e) U-Net #2.

Figure 9

Results of image processing after segmentation (large number of particles). (a) Input data; (b) Output data (ground truth). Results obtained using (c) conventional U-Net; (d) U-Net #1; (e) U-Net #2.

Figure 10

Training curve of (a) conventional U-Net, (b) improved U-Net #1, and (c) improved U-Net #2.

Figure 11

Results of image processing after segmentation (small number of particles). (a) Input data; (b) Output data (ground truth). Results obtained using (c) original ground truth data; and labeling data in which particles smaller than (d) 200 px; (e) 400 px; and (f) 600 px are removed.

Figure 12

Results of image processing after segmentation (large number of particles). (a) Input data; (b) Output data (ground truth). Results obtained using (c) original ground truth data; and labeling data in which particles smaller than (d) 200 px; (e) 400 px; and (f) 600 px are removed.