Image recognition method and apparatus, image acquisition method and apparatus, computer equipment, and non-volatile computer readable storage medium

Abstract

The invention discloses an image recognition method and apparatus based on a multilayer convolutional neural network, an image acquisition method and apparatus, computer equipment, and a non-volatilecomputer readable storage medium. According to the image recognition method and apparatus based on the multilayer convolutional neural network, the image acquisition method and apparatus, the computerequipment, and the non-volatile computer readable storage medium, a multilayer convolutional neural network model including three convolution layers and two pooling layers is constructed, the multilayer convolutional neural network model is trained for a training image of a first resolution by employing resolution normalization, the multilayer convolutional neural network model is tested for a test image of a second resolution by employing resolution normalization, the recognition of an image scene can be realized without using a full connecting layer, the complexity of a scene recognition algorithm is reduced, the calculating amount of scene recognition is small, and the time consumption of calculation is low.

Description

technical field [0001] The present invention relates to the technical field of image processing, in particular to an image recognition method based on a multi-layer convolutional neural network model, an image recognition device based on a multi-layer convolutional neural network model, an image acquisition method, an image acquisition device, a computer device, and Non-volatile computer readable storage medium. Background technique [0002] The existing methods of using artificially designed features to recognize image scenes have the disadvantages of long design cycle, poor robustness, and poor recognition ability for complex image scenes. The scene based on the convolutional neural network is that the recognition method needs to use a fully connected layer, which has the disadvantages of large amount of calculation and long calculation time. Contents of the invention [0003] Embodiments of the present invention provide an image recognition method based on a multi-laye...

AI Technical Summary

Problems solved by technology

[0002] The existing methods of using artificially designed features to identify image scenes have the disadvantages of long design cycle, poor robustness, and poor recognition ability for complex image scenes

The scene based on the convolutional neural network is that the recognition method needs to use a fully connected layer, which has the disadvantages of large amount of calculation and long calculation time.

Images