Unlock instant, AI-driven research and patent intelligence for your innovation.

A shopping guide behavior analysis method based on a lightweight multi-task convolution neural network

A convolutional neural network, behavior analysis technology, applied in the field of shopping guide behavior analysis, can solve the problem of no efficient solution, and achieve the effect of fast speed

Active Publication Date: 2019-03-29
ZHEJIANG UNIV OF TECH
View PDF4 Cites 10 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0005] There is currently no efficient solution for the rapid analysis of various attributes and behaviors of shopping guides

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • A shopping guide behavior analysis method based on a lightweight multi-task convolution neural network
  • A shopping guide behavior analysis method based on a lightweight multi-task convolution neural network
  • A shopping guide behavior analysis method based on a lightweight multi-task convolution neural network

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0021] A shopping guide behavior analysis method based on a lightweight multi-task convolutional neural network, comprising the following steps:

[0022] (1) cleaning data

[0023] Step 11: Remove images with a large number of mosaics in the training set;

[0024] This invention uses the data set of the BOT2018 New Retail Technology Challenge. For privacy protection, there are a large number of pedestrian images with mosaics in the data, which has a certain impact on the training of high-precision neural network models, so such images need to be removed. The invention designs a method of traversing pixel statistics in digital image processing technology to identify whether an image contains a large number of mosaics.

[0025] Each pixel value in a mosaic block is equal, and this feature is used to detect mosaic images. First, the pedestrian image is converted from a three-channel RGB image to a single-channel grayscale image, and then the grayscale image is divided into grid...

Embodiment 2

[0066] (1) Select experimental data

[0067] The present invention uses the data set of the BOT2018 New Retail Technology Challenge, the data is collected from real shopping mall scenes, and the shopping guides and customers in the images are images from the monitoring perspective. Divided into 5 scenes, a total of 5000 images, each image contains a different number of shopping guides and customers, the present invention divides these 5000 images into a training set and a test set at a ratio of 9:1, and extracts them on average.

[0068] Table 1 Dataset

[0069]

[0070] (2) Experimental results

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

PUM

No PUM Login to View More

Abstract

A shopping guide behavior analysis method based on a lightweight multi-task convolution neural network is proposed. Firstly, the data are cleaned to remove pictures which are not conducive to the convergence of neural network. Then, two kinds of Bottleneck module functions with and without residual structure are constructed. Based on these two kinds of Bottleneck, a lightweight multitask convolutional neural network is constructed. The training images are enhanced and normalized. The training model is pre-trained with large learning rate and then fine-tuned with small learning rate. Finally, lightweight multi-task convolutional neural network is used to recognize the behavior of shopping guide effectively and quickly.

Description

technical field [0001] The invention relates to a shopping guide behavior analysis method in the new retail field. Background technique [0002] With the deep integration of artificial intelligence and new retail, the use of deep learning technology to detect, analyze, and judge shopping guide attributes and analyze shopping guide behavior will help improve store operation efficiency and store management efficiency, optimize operational marketing strategies, increase sales conversion rates, Improving service efficiency and consumer experience will help the retail industry upgrade. [0003] In traditional retail scenarios, it is difficult to supervise shop assistants, and the activeness of shopping guides has an important impact on store sales. Due to the limited energy of managers and the uneven quality of shopping guides, the store manager cannot supervise all shopping guides at all times. Therefore, it often happens that shopping guides play with mobile phones during work...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

Application Information

Patent Timeline
no application Login to View More
Patent Type & Authority Applications(China)
IPC IPC(8): G06Q30/02G06N3/04G06N3/08
CPCG06N3/08G06Q30/0201G06N3/045
Inventor 赵云波林建武李灏宣琦
Owner ZHEJIANG UNIV OF TECH