基于优化AlexNet的花卉识别

Flower recognition based on optimized alexnet

准确并高效地识别花卉在自动化种植、机械采摘、病虫害防治、鲜花定级等方面均具有重要意义。为了使手机等嵌入式设备对花卉识别更具有适配性,在AlexNet的基础上,利用深度可分离卷积减少运算量,调整自适应池化层和全连接层减少参数量,将低维特征与高维特征相融合提升特征提取能力。实验结果表明优化后的模型可以有效降低运算量及参数量。模型体积仅14.34 MB,浮点运算量减少53%,识别速度提高31%,充分验证了优化方法的有效性。

Accurate and efficient identification of flowers is of great significance in automatic planting,mechanical picking,pest control,flower grading and so on.In order to make embedded devices such as mobile phones more adaptable to flower recognition,based on Alexnet,uses deep separable convolution to reduce the amount of computation,adjusts the adaptive pooling layer and full connection layer to reduce the amount of parameters,and combines low-dimensional features with high-dimensional features to improve the feature extraction ability.The experimental results show that the optimized model can effectively reduce the amount of computation and parameters.The volume of the model is only 14.34 MB,the floating-point operation is reduced by 53%,and the recognition speed is increased by 31%,which fully verifies the effectiveness of the optimization method.