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

A point location distribution-based commodity training set refinement acquisition method

A collection method and refined technology, applied in the computer field, can solve problems such as long time and manpower, poor model robustness, insufficient data sets, etc., to achieve the effect of reducing manual repeated operations, improving accuracy, and enhancing stability

Inactive Publication Date: 2019-06-18
上海小萌科技有限公司
View PDF0 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

In fact, there are certain problems in the above two methods. The first method will lead to insufficient data sets, which will make the trained model less robust and affect the actual application ability of the entire smart retail solution.
The second approach is similar to finding a needle in a haystack, which will consume a lot of time and manpower and affect the progress of the project

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 point location distribution-based commodity training set refinement acquisition method
  • A point location distribution-based commodity training set refinement acquisition method
  • A point location distribution-based commodity training set refinement acquisition method

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0025] Example 1: See Figure 1-7 , a refined collection method for product training sets based on point distribution, including the following steps:

[0026] A. Determine the commodity placement method; according to the actual situation, when collecting commodity image data, we determined three placement methods: single product, operation, and mixed. Assuming that there are four kinds of commodities (A, B, C, D) placed in the cabinet, each commodity is placed in columns, and each column is placed with 5 items. At this time, a 5×4 point distribution map is formed in the cabinet. When we collect training data, the complete picture of the three pendulum methods is as follows figure 1 , Figure 6 and Figure 7 as shown, figure 1 Use commodity A to illustrate the single-item arrangement, and the single-item arrangement of B, C, and D is the same. The three commodity placement methods have their own functions: (1) Single product: ensure that the commodity appears at all points...

Embodiment 2

[0044] Embodiment 2, on the basis of embodiment 1, the collection legend and the method of quantity of mixed pendulum method are as follows;

[0045] The types of goods in the mixed placement method are the same as those placed in the operation. The mixed placement method is more complicated in point dependence, but the way of placing the goods is easier. There are no strict requirements for the rows, columns and product categories in the point map. , as long as the 5×4 dots are filled. Assuming that the operation pendulum method of the positional relationship of ABCD is determined, the mixed pendulum method needs to combine the five states of "A", "B", "C", "D" and "empty". The image acquisition automatic optimization process is as follows:

[0046] Step 1: Set the number of dependencies t, and set the commodity placement symbols "A", "B", "C", "D" and "empty";

[0047] Step Two: Randomly figure 2 Place "A", "B", "C", "D" and "empty" on the 20 points of the 5×4 point map ...

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

The invention discloses a point location distribution-based commodity training set refined collection method. The method comprises the following steps of: A, determining a commodity placement method;B, determining a point location distribution relation; C, training image collection legends and number are determined according to the point location relation, by using the data collection scheme described by the invention, on one hand, manual repeated operation can be greatly reduced, and on the other hand, the accuracy of the algorithm can be improved, and the stability of the intelligent retailcabinet is enhanced.

Description

technical field [0001] The invention relates to the field of computers, in particular to a method for refined collection of commodity training sets based on point distribution. Background technique [0002] In smart retail solutions using computer vision technology, it is usually necessary to collect a large number of product data sets to train algorithm models. However, the image data set required by smart retail is different from the data used by conventional computer vision technology. The existing unmanned retail solutions based on computer vision and deep learning technology need to consider products from the perspective of products, commodities and external conditions. The collection of training sets is used to ensure the accuracy of the model, thereby improving the feasibility of the entire solution. [0003] There are generally two extreme methods in the existing commodity image data collection schemes. The first method usually only collects commodity data in a fix...

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): G06Q30/06G06K9/00
Inventor 龚飞
Owner 上海小萌科技有限公司