Collecting information for Technical Support#
To efficiently resolve a problem with LUNA SDK, collect all necessary information based on the error type and provide it to VisionLabs Technical Support. Possible error types include:
- Specific error
- Non-specific error
- Unexpected result
Contact Technical Support#
You can contact our Technical Support in either of the following ways:
- Via email: support@visionlabs.ai
- Via Service Portal: https://jira.visionlabs.ru/servicedesk/customer/portal/2
Specific error#
These errors usually occur when LUNA SDK is used incorrectly. Examples include:
- An estimator or detector does not work, resulting in an error when creating or using it.
- An error occurs when launching on a GPU device.
- A license error is received.
In such cases, study the full launch logs and understand what was launched and where.
To get detailed logging in LUNA SDK, follow these steps:
1․ In the luna-sdk/data/runtime.conf configuration file, set the verboseLogging
parameter to 4.
<param name="verboseLogging" type="Value::Int1" x="4" />
2․ In the luna-sdk/data/faceengine.conf configuration file, set the verboseLogging
parameter to 4.
<param name="verboseLogging" type="Value::Int1" x="4" />
3․ In the luna-sdk/data/trackengine.conf configuration file, set the severity
parameter to 0.
<param name="severity" type="Value::Int1" x="0" />
If you know which module the error occurs in, provide only that module's log by changing the value only in the relevant configuration file. If unsure, collect all logs.
Non-specific error#
Examples of non-specific errors include:
- An application crashes at an uncertain time.
- An application freezes unexpectedly.
- There is a memory leak.
In such cases, you need to understand in detail the application operation scenario, including what is called and in what sequence.
Provide the following information:
- The exact version of LUNA SDK (e.g., v.5.22.2, build for CentOS 8).
- Information about the environment where the application runs (e.g., Docker container, launch via Python bindings).
- Full launch logs.
- Additional information like crash dumps, reports from third-party utilities, and system logs.
- Code reproducing the problem, if any.
Unexpected Result#
Unexpected results may occur due to:
- Incorrect use of LUNA SDK
- Algorithm errors
- Launching in unexpected conditions
Examples include:
- A face is present in a photo or video, but the detector doesn't see it.
- A person is smiling, but the emotion estimator indicates sadness.
Reasons for unexpected results vary, such as:
- Incorrect use of LUNA SDK, for example, a wrong threshold in a configuration file.
- Incorrect input data, such as a poor-quality video or heavily compressed frames.
- Occasional algorithm errors.
- New data for the algorithm.
To understand and address the issue, provide:
- Full launch logs.
- All configuration files used during the launch:
- luna-sdk/data/runtime.conf
- luna-sdk/data/faceengine.conf
- luna-sdk/data/trackengine.conf
- An estimate of how often the unexpected result occurs, for example, every frame or once in a thousand frames.
- Examples of data that produce unexpected results.