Artificial intelligence system development method based on graphical mode

Abstract

The invention relates to an artificial intelligence system development method based on a graphical mode, and belongs to the technical field of software research and development and artificial intelligence. An artificial intelligence algorithm is divided into building block modules, a visual programming interface is provided in a mode of combining a webpage and a client, and a user can construct an artificial intelligence system in a mode of dragging building blocks without directly editing codes. The development of the system becomes visual and simple, and the technical threshold is reduced. According to the method, different types of algorithm codes are converted into specific code templates, the integration of artificial intelligence algorithms is realized in a form of modifying specific parameters in the templates, and a user can directly use the integrated building block modules to increase the functions and performance of the system. By providing wide algorithm choices, a user can quickly construct an artificial intelligence system with advanced functions, and the algorithm does not need to be developed and optimized by himself / herself. According to the method, the development efficiency of the artificial intelligence system is improved, and the technical threshold of development of the artificial intelligence system is reduced.

Description

Technical Field

[0001] The present invention relates to an artificial intelligence system development method based on a graphical approach, and belongs to the technical field of software development and artificial intelligence. Background Art

[0002] At present, graphical programming has attracted much attention. This method allows ordinary users to implement a system that meets their own needs according to their own application requirements without editing code, which lowers the development threshold and promotes the popularization of computer technology.

[0003] Graphical programming is a standardized way of communication that uses a program with an interface to define computer programs and issue instructions to the computer in a non-code way. For example, in the field of education, graphical programming can help students better understand the execution process and logical structure of the program, and cultivate their computational thinking and creativity. In the field of scientific research, graphical programming can improve the work efficiency of researchers, lower the threshold of programming, and enable more people to participate in scientific research. In the creative field, graphical programming can help creators turn their imagination into reality and achieve more complex interactive effects and visual effects.

[0004] In recent years, more and more graphical programming software has gained a good market. For example, Scratch, written based on HTML5, can help young people learn programming and express their wishes more intuitively and vividly during the creative process. The Mind+ platform directly supports the mainstream open source hardware commonly used in primary and secondary school maker education, and can program and control hundreds of commonly used hardware modules. KNIME contains various types of graphical modules, thousands of modules and hundreds of examples that can be run directly.

[0005] However, none of the existing graphical programming software involves artificial intelligence algorithms or componentizes artificial intelligence algorithms in a suitable manner, which is insufficient to support users' learning of artificial intelligence. Summary of the invention

[0006] The purpose of the present invention is to creatively propose a method for developing an artificial intelligence system based on a graphical approach in order to effectively solve the problems of high threshold for developing intelligent systems and other problems in view of the defects and deficiencies in the prior art. This method effectively divides the artificial intelligence algorithm into building block modules. Users can develop artificial intelligence systems without directly editing the code, which simplifies the development process of artificial intelligence systems. Through the visual programming interface and the building block system construction method, the development efficiency of the artificial intelligence system is greatly improved.

[0007] The innovative features of the present invention include:

[0008] 1. How to create a visual programming interface: The present invention provides a visual programming interface in the form of a web page and a client. Users can build an artificial intelligence system by dragging building blocks without directly editing the code. This makes system development intuitive and simple, and reduces the technical threshold.

[0009] 2. How to integrate artificial intelligence algorithms: After sorting out the basic artificial intelligence algorithm modules, the present invention converts different types of algorithm codes into specific code templates, and implements the integration of artificial intelligence algorithms by modifying specific parameters in the templates. Users can directly use these integrated building blocks to increase the functionality and performance of the system. By providing a wide range of algorithm options, users can quickly build artificial intelligence systems with advanced functions without having to develop and optimize algorithms themselves.

[0010] Beneficial Effects

[0011] Compared with the prior art, the method of the present invention has the following advantages:

[0012] 1. Users do not need to write complex codes directly, but can build systems through simple drag and drop and combination operations. This enables developers to build prototypes and iterate development faster, and speed up the launch time of artificial intelligence systems, greatly improving the development efficiency of artificial intelligence systems.

[0013] 2. Ordinary users can create customized AI systems without in-depth programming knowledge or algorithm background, which lowers the technical threshold for AI system development.

[0014] 3. The building block system construction method enables users to freely select and combine different building blocks to build artificial intelligence systems. This flexibility allows users to customize the functions and performance of the system according to specific needs, meet the personalized needs of different industries and users, and provide more room for creativity and innovation. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 It is a login interface design diagram in the embodiment.

[0016] Figure 2 It is a design diagram of the project management interface in the embodiment.

[0017] Figure 3 2. It is a design drawing of the workbench in the embodiment.

[0018] Figure 4 It is a schematic diagram of the system architecture of the present invention.

[0019] Figure 5 It is a diagram showing an example of a construction project in the embodiment.

[0020] Figure 6 2 is an example diagram of building blocks in the embodiment.

[0021] Figure 7 is an example diagram of an input data set path in the embodiment.

[0022] Figure 8 2 is a diagram showing an example of a training result in an embodiment.

[0023] Fig. 9 4 is a diagram showing an example of displaying training error information in the embodiment. DETAILED DESCRIPTION

[0024] The method of the present invention is described in detail below with reference to the accompanying drawings and embodiments.

[0025] The method of the present invention is implemented by adopting the following technical solutions.

[0026] A graphically based artificial intelligence system development method, including web page end, back end (web page background, usually deployed on the server), server, and client. Figure 4 , the design process includes the following steps:

[0027] Step 1: Block design.

[0028] Neural networks based on artificial intelligence involve various building blocks, including image preprocessing, data input, neural networks, learning methods, evaluation functions, Alphabeta pruning, audio to spectrum, gesture recognition, face recognition, chessboard creation, evolutionary computing, fitness calculation, SSH connection, speech recognition, word vector conversion, similarity calculation, and emotion setting.

[0029] These building blocks can modify specific internal parameters to achieve their corresponding functions.

[0030] Step 2: After receiving the user request, the web page sends the request signal containing user information, request type, and project information to the backend via the HTTP protocol.

[0031] Step 3: After receiving the request, the backend analyzes the request and performs corresponding operations. For training, running, verification and other requests, the required information such as user name, project number, data path, etc. is sent to the server via HTTP protocol, and the corresponding data in the database is modified.

[0032] For the creation, deletion, etc. of items, modify the data in the database.

[0033] Step 4: After receiving the data from the backend, the server generates and saves the corresponding data files for the training, running, and verification requests. Then, the server packages the required user name, data path, project number, and data file together and sends them to the client.

[0034] Step 5: After the client receives the data packet and saves it locally, it first decompresses the package, reads the data file, determines the type of artificial intelligence model, and modifies the corresponding parameters to build the corresponding artificial intelligence model locally. Subsequently, the model is trained, run, or verified according to the parameters given by the user. After completing the operation, the client sends the generated result path, result information, or error information to the server.

[0035] Step 6: After receiving the result packet from the client, the server sends the data to the backend.

[0036] Step 7: The backend records the information in the database based on the result information.

[0037] Step 8: After the front end detects that the task is completed, it sends a request to the back end to obtain the result.

[0038] Step 9: The backend reads the corresponding result information in the database and returns the result.

[0039] Step 10: The front end displays the results returned by the back end on the web page based on the type and whether it is successful.

[0040] Example

[0041] This embodiment describes the use scenario and specific use process of the present invention to demonstrate the functions and advantages of this method.

[0042] The following is a specific usage scenario:

[0043] An educational institution is provided that wishes to provide students with an easy-to-use platform that enables them to create and train artificial intelligence models graphically without having to edit code directly. Using the method of the present invention, the educational institution can easily achieve this goal.

[0044] First, the educational institution creates a project on the user front end and uploads the relevant data set. The administrator role can access the project management function and assign corresponding permissions to teachers and students. Teachers can access the project management and document management functions to upload course materials and assign them to students. Students can use the project management and document management functions to learn and practice related courses.

[0045] Through the intuitive interface design and simple interaction of the user front end, students can select and drag different building blocks to build their own artificial intelligence models. They can select the input layer, hidden layer and output layer of the model, and set the corresponding parameters and connections. Through the graphical interface, students can intuitively understand the structure and function of the model without having to delve into the programming details.

[0046] Once students have completed model building, they can choose to train the model. With the support of the code generation module, the software will generate the corresponding code based on the settings of the students on the graphical interface and send it to the code execution module. The code execution module runs the students' code locally and provides real-time feedback on the training process, including changes in the loss function and improvements in model accuracy.

[0047] Through the visual training results provided by the software, students can intuitively observe the performance of the model during training and make adjustments and improvements based on feedback. The software can also provide prompts and error feedback to help students correct mistakes and deepen their understanding of artificial intelligence.

[0048] The following are the specific steps for a student without any programming knowledge to use the educational platform developed based on this method:

[0049] Step 1: Open the platform, open the client, and open the browser, go to the corresponding website and log in, such as Figure 1 .

[0050] Step 2: Build the project and enter the project management interface, such as Figure 2 . Create a new project and select machine learning, such as Figure 5 . Enter the project workbench interface, such as Figure 3 . Build the blocks required for face recognition, and select the data input block, face recognition block, image preprocessing block, neural network block, and learning method block in turn, such as Figure 6 Then set the parameters of each block; data input block: modify the number of categories and select the data type; image preprocessing block: set the number of batches of data to learn each time and the proportion of the validation set; neural network block: set the neural network model, set the fully connected layer, set the activation function, input dimension, and output dimension; learning method block: set the learning rate and number of training times, etc.

[0051] Step 3: Start training. Click the "Train" button, copy the path of the dataset and paste it into the pop-up dialog box. Figure 7 After the training is completed, the building block platform will display the curve of accuracy and loss, such as Figure 8 When a training error occurs, an error message will pop up on the right, such as Fig. 9 .

[0052] Step 4: Run the project, click "Run", and enter the image path to be tested in the pop-up dialog box. The operation method and result display of running the project are similar to "Training". Its purpose is to call the trained model to infer the data.

[0053] Step 5: Verify the model, click "Verify", and enter the verification folder path in the pop-up dialog box. The verification project is similar to "Training". The trained model can be verified to obtain the accuracy of the model on the required verification set. The result display method is the same as "Training" and "Running".

[0054] This embodiment demonstrates the application of the software of the present invention in the field of education. By developing artificial intelligence models in a graphical way, students can quickly get started and understand the basic concepts and operation methods of artificial intelligence without writing complex codes. This method provides a lower technical threshold, enabling more people to participate in the study and practice of artificial intelligence.

Claims

1. A graphically-based artificial intelligence system development method, comprising a web page, a back end, a server, and a client, characterized in that: The following steps are involved: Step 1: Block design; Neural networks based on artificial intelligence involve various building blocks, including image preprocessing, data input, neural networks, learning methods, evaluation functions, Alphabeta pruning, audio to spectrum, gesture recognition, face recognition, chessboard creation, evolutionary computing, fitness calculation, SSH connection, speech recognition, word vector conversion, similarity calculation, and emotion settings; These building blocks can modify specific internal parameters to achieve their corresponding functions; Step 2: After receiving the user request, the web page sends the request signal containing user information, request type, and project information to the backend via HTTP protocol; Step 3: After receiving the request, the backend analyzes the request and performs corresponding operations. For training, running, and verification requests, the required information, including user name, project number, and data path, is sent to the server via HTTP protocol, and the corresponding data in the database is modified. For the creation and deletion of projects, modify the data in the database; Step 4: After receiving the data from the backend, the server generates corresponding data files for the training, running, and verification requests and saves them; Then, the server packages the required user name, data path, project number, and data file together and sends them to the client; Step 5: After the client receives the data packet and saves it locally, it first decompresses the packet, determines the type of the artificial intelligence model by reading the data file, and modifies the corresponding parameters, thereby building the corresponding artificial intelligence model locally; Subsequently, the model is trained, run or validated according to the parameters given by the user; After completing the operation, the client sends the generated result path, result information or error information to the server; Step 6: After receiving the result packet from the client, the server sends the data to the backend; Step 7: The backend records the information in the database based on the result information; Step 8: After the front end detects that the task is completed, it sends a request to the back end to obtain the result; Step 9: The backend reads the corresponding result information in the database and returns the result; Step 10: The front end displays the results returned by the back end on the web page according to the type and whether it is successful.

Images