基于MobileNet和迁移学习的微带青烟叶图像识别

Greenish Tobacco Recognition Based on Transfer Learning and MobileNet

为了准确地挑选出微带青烟叶,提出了基于MobileNet和迁移学习的微带青烟叶图像识别方法.首先收集烟叶图像建立样本数据集;其次,对在ImageNet数据集上训练好的MobileNet进行微调,以使其适应烟叶图像识别;最后,基于迁移学习方法利用烟草样本训练集对微调后的MoblieNet模型进行训练,从而准确识别微带青烟叶.

In order to select the greenish tobacco more accurately,a method of greenish tobacco leaf image recognition based on MobileNet and transfer learning was proposed.Firstly,the tobacco images were collected and the dataset was established.Secondly,the MobileNet trained on the ImageNet dataset was fine tuned adaptively for tobacco leaf image recognition.Finally,the fine tuned MobileNet model was trained by use of tobacco image training set based on transfer learning to realize image classification,so as to identify the greenish tobacco accurately.