一种新型卷积神经网络植物叶片识别方法

为提高植物叶片识别的准确率及减少计算代价,在Pytorch框架下提出一种融合了深度卷积生成式对抗网络(DCGAN)和迁移学习(TL)的新型卷积神经网络叶片识别方法。首先,对植物叶片图像进行预处理,通过DCGAN对样本数据库扩充;其次,利用迁移学习将Inception v3模型应用于图像数据处理上,以提高植物叶片识别的准确率;最后,通过对比实验对该方法的有效性进行验证。结果表明:该方法可以获得96.57%的植物叶片识别精度,同时参数训练的迭代次数由4000次缩短到560次。

Plant Leaf Recognition Method With New Convolution Neural Network

In order to improve the recognition accuracy and reduce computing costs of plant leaf images,we proposed a new deep convolutional neural network recognition method under the framework of Pythorch,combining Deep Convolutional Generative Adversarial Networks(DCGAN)learning algorithm with Transfer Learning(TL).Firstly,the plant leaf image was preprocessed.The database was expanded by DCGAN.Secondly,the Inception v3 model was applied to the image data processing by migration learning to improve the plant leaf recognition accuracy.Finally,the effectiveness of the proposed method was verified by comparative experiments.The results show that the proposed method can obtain 96.57%plant leaf recognition accuracy,and the number of iterations of parameter training is shortened from 4000 to 560.