Training method of mask detection model, mask detection method and related equipment

Abstract

The invention discloses a training method of a mask detection model and a mask detection method. The calculation detection efficiency is improved. The method provided by the invention comprises the following steps: acquiring a training image; inputting the training image into the initial mask detection model, and performing feature extraction on the training image to obtain training image features; performing pruning sensitivity analysis on each to-be-pruned layer based on a self-adaptive normalized sensitivity calculation mode, and determining a pruning proportion corresponding to each to-be-pruned layer; for each to-be-pruned layer, pruning the to-be-pruned layer based on the pruning proportion to obtain pruned image features, and updating the mask detection model according to the pruned image features; performing precision evaluation on the mask detection model to obtain an evaluation value; and if the evaluation value is lower than a preset evaluation value, returning to the step of obtaining the training image for continuous execution, and if the evaluation value is not lower than the preset evaluation value, taking the obtained mask detection model as a trained mask detection model.

Description

technical field [0001] The invention relates to the field of image recognition, in particular to a training method for a mask detection model, a mask detection method, a device, a computer device and a storage medium. Background technique [0002] Under the global epidemic, the demand for detection systems for whether people wear masks in public places has risen sharply. In order to control the cost of the mask detection system, it is an economical way to deploy the mask detection system on the edge board. For example, Baidu Edgeboard FZ3B is an edge computing board based on FPGA architecture for embedded and edge deployment. [0003] In recent years, most target detection algorithms use deep learning methods, and such methods have shown a good detection performance, and the most commonly used method is the convolutional neural network. In order to enhance the nonlinear fitting ability of the convolutional neural network, the convolutional neural network is designed to be d...

AI Technical Summary

Problems solved by technology

In order to enhance the nonlinear fitting ability of the convolutional neural network, the convolutional neural network is designed to be deeper and wider, which makes the convolutional neural network very dependent on computing power and storage space, although the convolutional neural network uses a multi-layer neural network The advantages of image locality and image locality reduce a large number of parameters, but the amount of calculation is still huge, which makes the convolutional neural network model not suitable for most devices with limited computing resources.

[0004] Therefore, in the existing method, when the convolutional neural network model is applied to devices with limited resources, there is a problem of low computational and detection efficiency.

Images