Based on the Improved AlexNet Flower Identification System

In modern society, image recognition must identify not only images that are consistent with the template, but also images that are inconsistent with the template. This paper describes a method for classifying flowers, improving the structure of the traditional AlexNet network model through data augmentation, and changing the original convolution kernel to a deep separable convolution. The recognition accuracy of the AlexNet network model adopted in this paper is better than that of the original AlexNet network model, and the calculation rate is also greatly improved, which meets the expected requirements.