Rapid automatic waveguide recognition using YOLO and OpenCV for 3D waveguide fabrication

Abstract

3D waveguide is attractive because it has a potential that the waveguides are arranged, not only in plane, into three-dimension. Mosquito method, of which the core and cladding material sources are liquid and will be cured using UV-light, is known as one of the ways to fabricate the 3D waveguide. In this method, if multiple waveguides are fabricated, time until UV-cure is different between first waveguide and last one and it leads to core position shift because of gravity. To solve the problem, we considered to reduce the time from the first waveguide formation to the UV-cure using automatic waveguide fabrication with a camera. For the automatic waveguide fabrication, YOLO (you look only once) and OpenCV is exploited as machine-learning and image-processor, respectively. As a result, waveguide recognition time of 10 ms and image processing time of 50 ms, totally 60 ms, are realized.

Design and Control of an Ultra-Low-Cost Logistic Delivery Fixed-Wing UAV

“…The exported precision and recall curves after the training completion are depicted in Figure 25. Recall is the ratio of correctly detected targets to the total actual target count, with a higher recall rate signifying more accurate target detection [68,69]. From Figure 25, it can be deduced that the model developed in this study achieves a precision of 90% and a recall of 99.5%, indicating superior performance in target detection tasks.…”

“…The exported precision and recall curves after the training completion are depicted in Figure 25. Recall is the ratio of correctly detected targets to the total actual target count, with a higher recall rate signifying more accurate target detection [68,69].…”

The Linux based body falls alarm system

“…As long as you open this application, Qt calls the Opencv library to drive the camera to collect data, and then codes and processes the data through Opencv to determine if the object has fallen down for a long time, and at the same time, Deliver pictures to the remote in-depth learning side for learning. If the object is found to be in a fallen state, an on-board Wifi will immediately send an alert to the remote Pc and Phone sides, and the relevant personnel will receive timely information for the next urgent action [6]. In terms of deep learning, this paper adopts the YOLOv5 object detection algorithm and selects the YOLOv5s model with a simple structure.…”