基于深度学习的蝴蝶识别研究与实现

Research and implementation of butterfly recognition based on deep learning

科学技术的不断进步,促进了社会的发展,人们开始逐渐重视自然生态中的物种多样性,对各类物种的识别需求也日益增多。蝴蝶是自然界中与人类生活、生产密切相关的物种,因此,对蝴蝶的识别研究具有重要意义和价值。但目前的蝴蝶检测大多效率低下、检测成本高、检测准确率低,鉴于此,设计使用labelme作为数据集的标注工具,使用深度学习框架PyTorch作为基本算法库,在PyCharm的集成开发环境上编码实现YOLOv5目标检测模型并进行训练,利用Qt Designer设计可视化界面,再使用PyQt5完成可视化界面并实现调用模型进行检测的功能,最后将检测结果与Mask RCNN、SSD算法的检测结果进行对比,对比结果显示,本文使用的方法明显优于其它网络,最后,基于上述步骤完成对野外环境中蝴蝶的识别研究与实现。

The continuous progress of science and technology promotes the development of society.People begin to pay more and more attention to the diversity of species in natural ecology,and the need for identification of various species is also increasing.Butterfly is a species closely related to human life and production in nature,so it is of great significance and value to identify and study the butterfly.However,most of the current butterfly detection methods have low efficiency,high detection cost and low detection accuracy.In view of this,labelme is used as the annotation tool of data set,PyTorch,a deep learning framework,is used as the basic algorithm library,and the YOLOv5 target detection model is coded and trained on the integrated development environment of PyCharm.Qt Designer is used to design the visual interface,and PyQt5 is used to complete the visual interface and realize the function of calling the model for detection.Finally,the detection results are compared with those of Mask RCNN and SSD algorithms.The comparison results show that the method used in this paper is obviously superior to other networks.Based on the above steps,the identification of butterflies in the wild environment is studied and realized.