The data of the article“Automatic fabrication system of optical micro-/nanofiber based on deep learning”

This article utilizes image segmentation methods in computer vision to create a high-quality multi-scale micro/nanofiber dataset. The YOLOv8-FD algorithm based on small object detection improvement is used to automatically detect the diameter of micro/nanofibers. The system can achieve measurement and automated preparation for micro/nano fiber with a diameter of 462 nm − 125 μm within an error of 2.95%, and with the increase of fiber diameter, the error gradually decreases. The optical imaging resolution of a single pixel in the system is 65.97 nm, and the average detection time is 9.6 ms. This work is suitable for high-precision real-time measurement and automatic precise preparation of micro/nano fibers. Figure 1 shows the network structure of micro/nanofiber diameter detection based on deep learning. Figure 2 shows the automatic preparation system of micro/nanofiber based on deep learning. Figure 3 shows the original image of Loss and mAP changing with epoch during the training process, where the opju file can be opened using Origin software. Figure 4 shows the visualization and results of the deep learning model. Figure 5 shows the comparison of the segmentation results of the original YOLOv8 and YOLOv8-FD micro/nano fiber images. Figure 6 shows the AFM scanning image of the micro/nano fiber. The ibw file can be opened using Gwyddion.