Open AccessA Nuclear detection method Based on improved You Only Look Once v5s Framework
Author(s) -
Yingbo Zhang,
Jingcun Wang,
Feng Shen
Publication year - 2021
Publication title -
journal of physics. conference series
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.21
H-Index - 85
eISSN - 1742-6596
pISSN - 1742-6588
DOI - 10.1088/1742-6596/2010/1/012187
Subject(s) - computer science , residual , task (project management) , artificial intelligence , pattern recognition (psychology) , carry (investment) , data mining , machine learning , algorithm , management , finance , economics
Extracting human cell smear samples through automated scanners, using target detection algorithm to detect its nucleus is an important content in cell detection. Due to the different shapes, different sizes and complex backgrounds of nuclei in the samples, automatic detection of nuclei is a challenging task. In order to solve the problem of complex small target recognition in nuclear detection and improve detection accuracy, based on the YOLOV5S framework of one-stage target detection, make certain improvement, add new residual module to the original network Neck layer, at the same time carry on network widening operation to the subsequent framework. The algorithm visualizes the results of training and testing through Tensorboard. Experimental results show that the accuracy of the improved YOLOV5S network is 1.6% and 4.4% higher than that of the original YOLOV5S network and YOLOV3 network. the improved network has strong real-time performance and high accuracy, which can meet the actual use requirements.