Production quality prediction method and system based on edge computing gateway

Abstract

The embodiment of the invention provides a production quality prediction method and system based on an edge computing gateway, after a sample production sampling characteristic data sequence uploaded by the edge computing gateway is detected, the sample production sampling characteristic data sequence is disassembled to generate a target number of sample production sampling characteristic data clusters; inputting one sample production sampling feature data cluster into the initialized neural network to obtain production quality prediction data, determining training error values of all samples in the sample production sampling feature data cluster according to the production quality prediction data and a set training error function, and performing parameter updating on the neural network according to the training error values of all the samples, and performing multiple rounds of traversal updating to generate an updated neural network. And if the next sample production sampling feature data cluster exists, using the updated neural network as an initialized neural network and repeating the process, otherwise, using the updated neural network as a target production quality prediction network. The method improves the accuracy and efficiency of production quality prediction.

Description

Technical Field

[0001] The present invention relates to the field of edge computing technology, and in particular to a production quality prediction method and system based on an edge computing gateway. Background Art

[0002] Edge computing is a computing model that pushes computing tasks, services, and resources from central nodes to the edge of the network, allowing data to be processed at or near the source to improve response speed and reduce bandwidth utilization. However, current edge computing systems mainly focus on data collection and transmission, and do not fully utilize this data for real-time decision-making and prediction.

[0003] On the other hand, neural networks are a powerful machine learning tool that can perform deep learning on large amounts of complex data and then make decisions or predictions based on the learned knowledge. However, when the training data categories are uneven, neural networks may have overfitting or underfitting problems, affecting the accuracy of predictions.

[0004] At the same time, in a production environment, the prediction of production quality usually requires a large amount of historical data and a complex analysis process, which often leads to inefficient production quality prediction and the accuracy of the prediction results cannot be guaranteed.

[0005] Therefore, it is of great theoretical and practical value to develop a technology that can combine edge computing and neural networks, take into account the problem of uneven data categories, and improve the accuracy and efficiency of production quality prediction. Summary of the invention

[0006] In view of this, the purpose of an embodiment of the present invention is to provide a production quality prediction method and system based on an edge computing gateway. After detecting the sample production sampling feature data sequence uploaded by the edge computing gateway, it is disassembled to generate a target number of sample production sampling feature data clusters, and a sample production sampling feature data cluster is input into the initialization neural network to obtain production quality prediction data. The data cluster contains training data with uneven categories. The training error value of each sample in the sample production sampling feature data cluster is determined based on the production quality prediction data and the set training error function, and the neural network parameters are updated according to the training error value of each sample, so as to perform multiple rounds of traversal updates to generate an updated neural network. Finally, if there is a next sample production sampling feature data cluster, the updated neural network is used as the initialization neural network and the above process is repeated, otherwise the updated neural network is used as the target production quality prediction network. This method improves the accuracy and efficiency of production quality prediction.

[0007] According to one aspect of an embodiment of the present invention, a production quality prediction method and system based on an edge computing gateway are provided, the method comprising:

[0008] When a sample production sampling feature data sequence uploaded by the edge computing gateway is detected, the sample production sampling feature data sequence is disassembled to generate a target number of sample production sampling feature data clusters;

[0009] Inputting one of the sample production sampling feature data clusters into an initialized neural network to obtain production quality prediction data of the sample production sampling feature data cluster, wherein the sample production sampling feature data cluster contains training data with uneven categories;

[0010] Determining a training error value of each sample production sampling feature data in the sample production sampling feature data cluster according to the production quality prediction data and a set training error function, wherein the training error value can be updated according to a prediction accuracy index of the production quality prediction data;

[0011] The network parameters of the initialized neural network are updated according to the training error value of the sampled characteristic data produced by each sample, so as to perform multiple rounds of traversal updates on the initialized neural network to generate an updated neural network;

[0012] Detect whether there is a next sample production sampling feature data cluster. If there is a next sample production sampling feature data cluster, change the updated neural network to the initialized neural network and return to execute the step of inputting one of the sample production sampling feature data clusters into the initialized neural network to obtain the production quality prediction data of the sample production sampling feature data cluster. Otherwise, change the updated neural network to the target production quality prediction network.

[0013] In an alternative implementation, after the sample production sampling characteristic data sequence is disassembled to generate a target number of sample production sampling characteristic data clusters, the method further includes:

[0014] It is detected whether each of the sample production sampling characteristic data clusters meets the target condition, so as to obtain the sample production sampling characteristic data clusters that meet the target condition.

[0015] In an alternative implementation, the target condition includes setting a weight interval, and the detecting whether each sample production sampling feature data cluster meets the target condition includes:

[0016] Counting the number of positive learning features and the number of negative learning features in each sample production sampling feature data cluster according to the labeled production quality category of each sample production sampling feature data;

[0017] Obtaining a comparison weight value of the number of positive learning features and the number of negative learning features in each of the sample production sampling feature data clusters;

[0018] Detecting whether the comparison weight value of each sample production sampling feature data cluster belongs to a set weight range;

[0019] If the comparison weight value of the sample production sampling feature data cluster belongs to the set weight interval, it is determined that the sample production sampling feature data cluster meets the target condition.

[0020] In an alternative implementation, the determining the training error value of each sample production sampling feature data in the sample production sampling feature data cluster based on the production quality prediction data and a set training error function includes:

[0021] Obtaining a prediction accuracy index of the sample production sampling feature data cluster based on the production quality prediction data;

[0022] Detecting whether the labeled production quality category of each of the sample production sampling feature data matches the production quality data of the forward learning feature;

[0023] If the labeled production quality category of the sample production sampling feature data matches the production quality data of the forward learning feature, determining the training error value of the sample production sampling feature data according to the first error algorithm in the training error function and the prediction accuracy indicator;

[0024] If the labeled production quality category of the sample production sampling feature data does not match the production quality data of the forward learning feature, the training error value of the sample production sampling feature data is determined based on the second error algorithm in the training error function and the prediction accuracy indicator.

[0025] In an alternative implementation, before detecting whether the labeled production quality category of each sample production sampling feature data matches the production quality data of the forward learning feature, the method further includes:

[0026] The marked production quality category of each of the sample production sampling feature data is encoded to obtain the encoded feature value of each of the sample production sampling feature data.

[0027] In an alternative implementation, the updating of the network parameters of the initialized neural network according to the training error value of each sample production sampling feature data includes:

[0028] Determining the optimized value of each network weight parameter in the initialized neural network according to the training error value of each sample production sampling feature data;

[0029] Each network weight parameter is updated according to the optimized value of each network weight parameter to train the initialized neural network once.

[0030] In an alternative embodiment, the prediction accuracy indicator is the prediction accuracy mean, the prediction accuracy mean variance or the prediction accuracy standard deviation.

[0031] According to another aspect of an embodiment of the present invention, a production quality prediction method and system based on an edge computing gateway are provided, the system comprising:

[0032] A disassembly unit is used to disassemble the sample production sampling feature data sequence to generate a target number of sample production sampling feature data clusters when detecting the sample production sampling feature data sequence uploaded by the edge computing gateway.

[0033] An acquisition unit, used for inputting one of the sample production sampling feature data clusters into an initialized neural network to acquire production quality prediction data of the sample production sampling feature data cluster, wherein the sample production sampling feature data cluster contains training data with uneven categories;

[0034] A determination unit, configured to determine a training error value of each sample production sampling feature data in the sample production sampling feature data cluster according to the production quality prediction data and a set training error function, wherein the training error value can be updated according to a prediction accuracy index of the production quality prediction data;

[0035] An updating unit, used for updating the network parameters of the initialized neural network according to the training error value of each sample production sampling feature data, so as to perform multiple rounds of traversal updates on the initialized neural network to generate an updated neural network;

[0036] A detection unit is used to detect whether there is a next sample production sampling feature data cluster. If there is a next sample production sampling feature data cluster, the updated neural network is used as the initialized neural network and the step of inputting one of the sample production sampling feature data clusters into the initialized neural network to obtain the production quality prediction data of the sample production sampling feature data cluster is returned; otherwise, the updated neural network is used as the target production quality prediction network.

[0037] According to another aspect of an embodiment of the present invention, a server is provided, comprising: a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory communicate with each other through the communication bus; the memory is used to store a computer program; the processor is used to implement any of the steps of the production quality prediction method based on the edge computing gateway when executing the computer program.

[0038] According to another aspect of an embodiment of the present invention, a readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps of the above-mentioned production quality prediction method based on edge computing gateway can be executed.

[0039] In order to make the above-mentioned purposes, features and advantages of the embodiments of the present invention more obvious and easy to understand, the embodiments will be described in detail below in conjunction with the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments are briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.

[0041] Figure 1 A schematic diagram showing components of a server provided by an embodiment of the present invention is shown;

[0042] Figure 2 A schematic diagram of a process of a production quality prediction method based on an edge computing gateway provided in an embodiment of the present invention is shown;

[0043] Figure 3 A functional module block diagram of a production quality prediction system based on an edge computing gateway provided in an embodiment of the present invention is shown. DETAILED DESCRIPTION

[0044] In order to enable students in the technical field to better understand the scheme of the present invention, the technical scheme in the embodiment of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiment of the present invention. Obviously, the described embodiment is only a part of the embodiment of the present invention, not all the embodiments. According to the embodiment of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0045] The terms "first", "second", "third", etc. (if any) in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein, for example. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0046] Figure 1 An exemplary component diagram of a server 100 is shown. The server 100 may include one or more processors 104, such as one or more central processing units (CPUs), each of which may implement one or more hardware threads. The server 100 may also include any storage medium 106 for storing any kind of information such as code, settings, data, etc. Non-limitingly, for example, the storage medium 106 may include any one or more combinations of the following: any type of RAM, any type of ROM, flash memory device, hard disk, optical disk, etc. More generally, any storage medium may use any technology to store information. Further, any storage medium may provide volatile or non-volatile retention of information. Further, any storage medium may represent a fixed or removable component of the server 100. In one case, when the processor 104 executes an associated instruction stored in any storage medium or a combination of storage media, the server 100 may perform any operation of the associated instruction. The server 100 also includes one or more drive units 108 for interacting with any storage medium, such as a hard disk drive unit, an optical disk drive unit, etc.

[0047] The server 100 also includes input / output 110 (I / O) for receiving various inputs (via input unit 112) and for providing various outputs (via output unit 114). One specific output mechanism may include a presentation device 116 and an associated graphical user interface (GUI) 118. The server 100 may also include one or more network interfaces 120 for exchanging data with other devices via one or more communication units 122. One or more communication buses 124 couple the components described above together.

[0048] The communication unit 122 may be implemented in any manner, for example, through a local area network, a wide area network (e.g., the Internet), a point-to-point connection, etc., or any combination thereof. The communication unit 122 may include any combination of hardwired links, wireless links, routers, gateway functions, name servers 100, etc. governed by any protocol or combination of protocols.

[0049] Figure 2 The flowchart of the production quality prediction method and system based on edge computing gateway provided by the embodiment of the present invention is shown. The production quality prediction method and system based on edge computing gateway can be Figure 1 The server 100 shown in FIG. 1 is executed, and the detailed steps of the production quality prediction method based on the edge computing gateway are introduced as follows.

[0050] Step S110, when detecting a sample production sampling feature data sequence uploaded by the edge computing gateway, disassembling the sample production sampling feature data sequence to generate a target number of sample production sampling feature data clusters;

[0051] Step S120, inputting one of the sample production sampling feature data clusters into an initialized neural network to obtain production quality prediction data of the sample production sampling feature data cluster, wherein the sample production sampling feature data cluster contains training data with uneven categories;

[0052] Step S130, determining a training error value of each sample production sampling feature data in the sample production sampling feature data cluster according to the production quality prediction data and a set training error function, wherein the training error value can be updated according to a prediction accuracy index of the production quality prediction data;

[0053] Step S140, updating the network parameters of the initialized neural network according to the training error value of each sample production sampling feature data, so as to perform multiple rounds of traversal updates on the initialized neural network to generate an updated neural network;

[0054] Step S150, detect whether there is a next sample production sampling feature data cluster. If there is a next sample production sampling feature data cluster, the updated neural network is used as the initialized neural network and returns to execute the step of inputting one of the sample production sampling feature data clusters into the initialized neural network to obtain the production quality prediction data of the sample production sampling feature data cluster. Otherwise, the updated neural network is used as the target production quality prediction network.

[0055] Based on the above steps, after detecting the sample production sampling feature data sequence uploaded by the edge computing gateway, this embodiment disassembles it to generate the target number of sample production sampling feature data clusters, and inputs a sample production sampling feature data cluster into the initialized neural network to obtain production quality prediction data, and the data cluster contains training data with uneven categories. The training error value of each sample in the sample production sampling feature data cluster is determined based on the production quality prediction data and the set training error function, and the parameters of the neural network are updated according to the training error value of each sample, so as to perform multiple rounds of traversal updates to generate an updated neural network. Finally, if there is a next sample production sampling feature data cluster, the updated neural network is used as the initialized neural network and the above process is repeated, otherwise the updated neural network is used as the target production quality prediction network. This method improves the accuracy and efficiency of production quality prediction.

[0056] In an alternative implementation, after the sample production sampling characteristic data sequence is disassembled to generate a target number of sample production sampling characteristic data clusters, the method further includes:

[0057] It is detected whether each of the sample production sampling characteristic data clusters meets the target condition, so as to obtain the sample production sampling characteristic data clusters that meet the target condition.

[0058] In an alternative implementation, the target condition includes setting a weight interval, and the detecting whether each sample production sampling feature data cluster meets the target condition includes:

[0059] Counting the number of positive learning features and the number of negative learning features in each sample production sampling feature data cluster according to the labeled production quality category of each sample production sampling feature data;

[0060] Obtaining a comparison weight value of the number of positive learning features and the number of negative learning features in each of the sample production sampling feature data clusters;

[0061] Detecting whether the comparison weight value of each sample production sampling feature data cluster belongs to a set weight range;

[0062] If the comparison weight value of the sample production sampling feature data cluster belongs to the set weight interval, it is determined that the sample production sampling feature data cluster meets the target condition.

[0063] In an alternative implementation, the determining the training error value of each sample production sampling feature data in the sample production sampling feature data cluster based on the production quality prediction data and a set training error function includes:

[0064] Obtaining a prediction accuracy index of the sample production sampling feature data cluster based on the production quality prediction data;

[0065] Detecting whether the labeled production quality category of each of the sample production sampling feature data matches the production quality data of the forward learning feature;

[0066] If the labeled production quality category of the sample production sampling feature data matches the production quality data of the forward learning feature, determining the training error value of the sample production sampling feature data according to the first error algorithm in the training error function and the prediction accuracy indicator;

[0067] If the labeled production quality category of the sample production sampling feature data does not match the production quality data of the forward learning feature, the training error value of the sample production sampling feature data is determined based on the second error algorithm in the training error function and the prediction accuracy indicator.

[0068] In an alternative implementation, before detecting whether the labeled production quality category of each sample production sampling feature data matches the production quality data of the forward learning feature, the method further includes:

[0069] The marked production quality category of each of the sample production sampling feature data is encoded to obtain the encoded feature value of each of the sample production sampling feature data.

[0070] In an alternative implementation, the updating of the network parameters of the initialized neural network according to the training error value of each sample production sampling feature data includes:

[0071] Determining the optimized value of each network weight parameter in the initialized neural network according to the training error value of each sample production sampling feature data;

[0072] Each network weight parameter is updated according to the optimized value of each network weight parameter to train the initialized neural network once.

[0073] In an alternative embodiment, the prediction accuracy indicator is the prediction accuracy mean, the prediction accuracy mean variance or the prediction accuracy standard deviation.

[0074] Figure 3 The functional module diagram of the production quality prediction system 200 based on the edge computing gateway provided by the embodiment of the present invention is shown. The functions implemented by the production quality prediction system 200 based on the edge computing gateway can correspond to the steps performed by the above method. The production quality prediction system 200 based on the edge computing gateway can be understood as the above server 100, or the processor of the server 100, or can also be understood as a component independent of the above server 100 or the processor that implements the functions of the present invention under the control of the server 100, such as Figure 3As shown, the functions of each functional module of the production quality prediction system 200 based on the edge computing gateway are explained in detail below.

[0075] The disassembly unit 210 is used to disassemble the sample production sampling feature data sequence to generate a target number of sample production sampling feature data clusters when detecting the sample production sampling feature data sequence uploaded by the edge computing gateway.

[0076] An acquisition unit 220, configured to input one of the sample production sampling feature data clusters into an initialized neural network to acquire production quality prediction data of the sample production sampling feature data cluster, wherein the sample production sampling feature data cluster contains training data of uneven categories;

[0077] A determination unit 230, configured to determine a training error value of each sample production sampling feature data in the sample production sampling feature data cluster according to the production quality prediction data and a set training error function, wherein the training error value may be updated according to a prediction accuracy index of the production quality prediction data;

[0078] An updating unit 240 is used to update the network parameters of the initialized neural network according to the training error value of each sample production sampling feature data, so as to perform multiple rounds of traversal updates on the initialized neural network to generate an updated neural network;

[0079] The detection unit 250 is used to detect whether there is a next sample production sampling feature data cluster. If there is a next sample production sampling feature data cluster, the updated neural network is used as the initialized neural network and the step of inputting one of the sample production sampling feature data clusters into the initialized neural network to obtain the production quality prediction data of the sample production sampling feature data cluster is returned. Otherwise, the updated neural network is used as the target production quality prediction network.

[0080] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0081] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above and that the present invention can be implemented in other specific forms without departing from the spirit or essential features of the present invention. Therefore, the embodiments should be considered exemplary and non-restrictive in all respects, and the scope of the present invention is defined by the appended claims rather than the above description, and it is intended that all variations falling within the meaning and scope of the equivalent elements of the claims be included in the present invention.

Claims

1. A production quality prediction method based on edge computing gateway, characterized in that: The method comprises: When a sample production sampling feature data sequence uploaded by the edge computing gateway is detected, the sample production sampling feature data sequence is disassembled to generate a target number of sample production sampling feature data clusters; Inputting one of the sample production sampling feature data clusters into an initialized neural network to obtain production quality prediction data of the sample production sampling feature data cluster, wherein the sample production sampling feature data cluster contains training data with uneven categories; Determining a training error value of each sample production sampling feature data in the sample production sampling feature data cluster according to the production quality prediction data and a set training error function, wherein the training error value can be updated according to a prediction accuracy index of the production quality prediction data; The network parameters of the initialized neural network are updated according to the training error value of the sampled characteristic data produced by each sample, so as to perform multiple rounds of traversal updates on the initialized neural network to generate an updated neural network; Detect whether there is a next sample production sampling feature data cluster. If there is a next sample production sampling feature data cluster, use the updated neural network as the initialized neural network and return to execute the step of inputting one of the sample production sampling feature data clusters into the initialized neural network to obtain the production quality prediction data of the sample production sampling feature data cluster. Otherwise, use the updated neural network as the target production quality prediction network.

2. The production quality prediction method based on edge computing gateway according to claim 1 is characterized in that: After the sample production sampling characteristic data sequence is disassembled to generate a target number of sample production sampling characteristic data clusters, the method further includes: It is detected whether each of the sample production sampling characteristic data clusters meets the target condition, so as to obtain the sample production sampling characteristic data clusters that meet the target condition.

3. The production quality prediction method based on edge computing gateway according to claim 2 is characterized in that: The target condition includes setting a weight interval, and the detecting whether each sample production sampling feature data cluster meets the target condition includes: Counting the number of positive learning features and the number of negative learning features in each sample production sampling feature data cluster according to the labeled production quality category of each sample production sampling feature data; Obtaining a comparison weight value of the number of positive learning features and the number of negative learning features in each of the sample production sampling feature data clusters; Detecting whether the comparison weight value of each sample production sampling feature data cluster belongs to a set weight range; If the comparison weight value of the sample production sampling feature data cluster belongs to the set weight interval, it is determined that the sample production sampling feature data cluster meets the target condition.

4. The production quality prediction method based on edge computing gateway according to claim 1 is characterized in that: The determining of the training error value of each sample production sampling feature data in the sample production sampling feature data cluster according to the production quality prediction data and the set training error function includes: Obtaining a prediction accuracy index of the sample production sampling feature data cluster based on the production quality prediction data; Detecting whether the labeled production quality category of each of the sample production sampling feature data matches the production quality data of the forward learning feature; If the labeled production quality category of the sample production sampling feature data matches the production quality data of the forward learning feature, determining the training error value of the sample production sampling feature data according to the first error algorithm in the training error function and the prediction accuracy indicator; If the labeled production quality category of the sample production sampling feature data does not match the production quality data of the forward learning feature, the training error value of the sample production sampling feature data is determined based on the second error algorithm in the training error function and the prediction accuracy indicator.

5. The production quality prediction method based on edge computing gateway according to claim 4 is characterized in that: Before detecting whether the labeled production quality category of each sample production sampling feature data matches the production quality data of the forward learning feature, the method further includes: The marked production quality category of each of the sample production sampling feature data is encoded to obtain the encoded feature value of each of the sample production sampling feature data.

6. The production quality prediction method based on edge computing gateway according to claim 1 is characterized in that: The updating of the network parameters of the initialized neural network according to the training error value of the sampled characteristic data produced by each sample comprises: Determining the optimized value of each network weight parameter in the initialized neural network according to the training error value of each sample production sampling feature data; Each network weight parameter is updated according to the optimized value of each network weight parameter to train the initialized neural network once.

7. The production quality prediction method based on edge computing gateway according to claim 4 is characterized in that: The prediction accuracy indicator is the prediction accuracy mean, the prediction accuracy mean variance or the prediction accuracy standard deviation.

8. A production quality prediction system based on edge computing gateway, characterized in that: include: A disassembly unit is used to disassemble the sample production sampling feature data sequence to generate a target number of sample production sampling feature data clusters when detecting the sample production sampling feature data sequence uploaded by the edge computing gateway. An acquisition unit, used for inputting one of the sample production sampling feature data clusters into an initialized neural network to acquire production quality prediction data of the sample production sampling feature data cluster, wherein the sample production sampling feature data cluster contains training data with uneven categories; A determination unit, configured to determine a training error value of each sample production sampling feature data in the sample production sampling feature data cluster according to the production quality prediction data and a set training error function, wherein the training error value can be updated according to a prediction accuracy index of the production quality prediction data; An updating unit, used for updating the network parameters of the initialized neural network according to the training error value of each sample production sampling feature data, so as to perform multiple rounds of traversal updates on the initialized neural network to generate an updated neural network; A detection unit is used to detect whether there is a next sample production sampling feature data cluster. If there is a next sample production sampling feature data cluster, the updated neural network is used as the initialized neural network and the step of inputting one of the sample production sampling feature data clusters into the initialized neural network to obtain the production quality prediction data of the sample production sampling feature data cluster is returned; otherwise, the updated neural network is used as the target production quality prediction network.

9. A server, characterized in that: include: A processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory communicate with each other via the communication bus; The memory is used to store computer programs; the processor is used to implement the steps of the production quality prediction method based on edge computing gateway as described in any one of claims 1 to 7 when executing the computer program.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the production quality prediction method based on edge computing gateway described in any one of claims 1 to 7 are implemented.

Images