Embedded crowd density estimation method based on a convolutional neural network model

Abstract

The invention discloses an embedded crowd density estimation method based on a convolutional neural network model and the convolutional neural network model of embedded crowd density estimation. The model is used for realizing the method. The method comprises the following steps of: nesting structures of three convolutional neural branches with the output capability of generating a crowd density map; the model has three operation modes; and after the training image is preprocessed, training a convolutional neural network model, inputting the image to the trained convolutional neural network model, selecting one of the three operation modes, outputting a crowd density map corresponding to the selected mode, and performing an integral operation on the output density map to obtain the total number of people of the image. The convolutional neural network model is light in weight, the accuracy is higher than that of a convolutional neural network model of the same magnitude, three deployment modes can be switched at will, the speeds of the modes are different, and the speeds can be selected according to actual conditions.

Description

technical field [0001] The invention relates to the field of computer vision, in particular to an embedded crowd density estimation method based on a convolutional neural network model and a convolutional neural network model for embedded crowd density estimation. Background technique [0002] Relying on embedded real-time computing for localization can overcome the shortcomings of existing methods based on convolutional neural networks. Comparison of server-side GPU-based and embedded-based crowd density estimation methods image 3 As shown, in comparison, the latter is more convenient and easy to use and helps to save costs. However, when the model is applied to embedded devices, the performance of the device, the size of the image, and the real-time requirements of customers may change. It is impractical to design networks according to different situations. [0003] For this reason, we need to design a lightweight and speed-adjustable convolutional neural network model f...

AI Technical Summary

Problems solved by technology

However, when the model is applied to embedded devices, the performance of the device, the size of the imaging, and the real-time requirements of customers may change. It is impractical to design networks according to different situations.

Images