SDK

Summarization Problems

Q: Can the SDK support local detection? Or must need Internet connection?

The SDK support local detection, no need to connect to the internet. You only need to download the trained model files and SDK toolkit from the cloud platform to perform offline local detection.

Q: How to check if the configuration of the computer environment supports it?

You can reference environment configuration in the link below: ‍

AI Model Deployment / How To Use SDK(C++)

You can also run the detectEnvironmentAndPredict.exe file to check your environment configuration, see details in:

‍https://github.com/neurobot-ai/neurobot_sdk_demo/tree/main/C++/01.detectEnvironmentAndPredict

Development Operation Problems

The following italicized problems are the problems encountered by customers, please first focus on them.

Q: Error at loading model (load_model()function) when executing program.

Common error messages are as follows:

In the event of such a problem, please check that the following categories are set correctly:

(1) Is the model type correct

When select the CPU model, namely device_name = "CPU", ensure load_model function to load the model file for the CPU model. The CPU models folder contains the following files:

When the GPU model type is selected, i.e. device_name="cuda", make sure that the model file loaded by the load_model function is the GPU model. The GPU models folder contains the following files:

(2) The model reference path is the path to the model folder‍

Model package after decompression, you can get the model folder. In the example, model_path is the path to load the model, which needs to be the path to the models folder instead of a file.

Q: An exception occurred when running with the cpu model.

This is a problem that many people encounter when working with the CPU model.

There may be a conflict between the old version file with the same name in the system and the onnxruntime.dll file in the SDK package. When the system loads the dll file, the old version of the system will be loaded, this will result in an exception.

Solution:

Copy the onnxruntime.dll file in the bin directory of the SDK package to the folder where the "exe file" generated by the project is located, that is, "... \x64-release ". Double-click the generated "exe file" to check whether it works properly.

In some cases, double-clicking the generated exe file works fine, but still generates an error when running in VisualStudio. Please try to copy the onnxruntime.dll file in the SDK package to the C:\Windows\system32 path, overwriting the original version file (system permission is required).

Q: Why is the path error displayed when I run the program after modifying the model path and picture path in the code sample file?

Use a double backslash "\\" for path delimiters in code, not a single backslash "\". For example, the file path for storing pictures is "C:\Users\81098\Desktop\test_pic". If you put this in the code, an error will be reported!

Please use double backslashes like: "C:\\Users\\81098\\Desktop\\test_pic"

Q: How can I find "Solution Explorer" in my project?

Click on "View", then click on "Solution Explorer".

Q: When the program is compiling, a message is displayed indicating that some components cannot be found. The environment variable configuration is invalid.

One of the common reasons why a configured environment variable is invalid is that there may be a conflict with the originally configured environment variable, resulting in an abnormal error when calling the SDK. Please try:

• Delete conflicting environment variables in path.
• If you are using Windows 11, you need to move the added environment variable up to the top.

Q: How to call SDK for development by using CPU?

The steps for calling the SDK with cpu are the same as those for using an Nvidia graphics card. Just modify the variable device_name in the code and change "cuda" to "cpu". See also:

AI Model Deployment / How To Use SDK(C++) / 4.1 load_model()

Error 56: "No permission to write authorized files" is reported when running the SDK.

When the SDK is running, the program detects the authorization file 'license.txt'. This error is reported when the program does not have write permission for the file.

Method of viewing permissions:

• Find the file or folder you want to check permissions on, right click on it and select "Properties".
• First, on the General TAB of Properties, check whether the property Read-only is checked. If yes, deselect Read-only and try again to check whether the SDK runs properly. If the above operation does not work, try the subsequent method then:

• In the Properties window, switch to the Security TAB. This TAB displays the access permissions of the current user and group to the file.
• Click the "Edit" button to view or change permission Settings. If you need to change permissions, make sure you have sufficient administrator permissions to do so.
• In the Edit window, you can add or remove users, groups, or specific permissions, such as read, write, execute, etc. You can also change existing permission Settings.
• Click the "OK" button to save your changes. Please note that when you change permission Settings, it may affect other users or applications' access to the file. It is important to note that before making any permission changes, it is important to understand all the associated risks and consequences and to follow best practices and security policies. Incorrect permission Settings can lead to security vulnerabilities or data corruption, so handle with caution. If you are unsure how to set permissions for a file or folder, consult your system administrator or technical support.

If the error is still reported after the change:

The following problems may exist 1. The file may be in use by other software. 2. Insufficient disk space. 3. Editor permission error.

Q: Environment variables are too large when configuring the environment (as follow)