Skip to main content
Version: Next

Monitor Edges

Updated 2025.02.18

Users can monitor the connection status, inference status, and performance of Edges through the Edge Conductor.

Edge Status​

Displays the connection status and state between the Edge Conductor and Edges.

  • Disconnected: The WebSocket connection between the Edge App and Edge Conductor is disconnected
  • No Stream: The Edge App is connected to the Conductor, but there is no AI Solution or model information deployed on the App. Inference cannot be performed in this state.
  • Ready: The AI Solution and model are deployed, and the system is ready for inference.
  • Inferencing: The system is currently performing inference.

Edge Inference Score​

Edges → Inf. result

Users can set an alarm threshold for the inference score when creating a stream. If the score of the inference result received from the Edge is lower than the threshold, the inference is classified as a warning. When the inference result is uploaded to the Edge Conductor, the system checks the inference score and sends an alert if it is below the threshold. Users can check the score values of each inference through the Inference Result Table, click on the inference content to view detailed data about the inference, and if it is determined that the model has inferred incorrectly or the model's accuracy has decreased, they can improve the inference accuracy of the Edges by reconstructing the training data and redeploying the model.


Inference Result​

Edges → Inf. result

All AI Solutions of Mellerikat output summarized information about the inference results for performed inference tasks. The summarized information of the inference results is structured as follows:

Summary Information of Inference Results​

  • Time : The time the inference was performed
  • Input File : Input File of inference
  • Result : The result inferred by the model
  • Score: The performance metric of the model's inference result
  • Model Version: The version of the model
  • Note: Additional information or remarks about the inference
  • Output: Inference result detailed data. Table or image data
  • Log: Inference Log

The summarized information of the inference results is designed to output values suitable for the characteristics of the problem that the provider of the AI Solution intends to solve. For the settings of the AI Solution for the inference result, refer to the AI Solution. Additionally, by clicking on the result of each inference task, you can view detailed data, which supports image or table data.


Detailed Inference Score​

The score represents the model's performance or information that can be referenced for labeling and retraining the model. For example, in anomaly detection, performance can be represented by the number of anomaly points relative to the total number of points, or in classification problems, if the score is a probability value, data where the model predicted with uncertainty can be checked.


Inference Result Detail​

Edges → Inf. result

View as Album​

Inference results for image data can be viewed as an album. The album provides a probability bar graph of the class inferred by the model at the bottom of the image, allowing intuitive verification of the model's performance.

View as Table​

Displays the tabular data used for inference in a table. Users can view the data used for inference through the screen.


Inference failure​

Edges → Inf. Failure

In the case of inference failure, provide information to help the user identify the cause of the problem.

Summary Information​

  • Time : The time the inference was performed
  • Input File : Input File of inference
  • Model Version: The version of the model
  • Message : The messge of Inference Failure
  • Log : Inference Failure Log

Providing Standard Inference Results for Customizable User Analytics​

Mellerikat is a service that can solve various problems with various AI Solutions. Therefore, there are various forms of inference results, and to verify inference results, individual visualization methods for inference results for each AI Solution or problem are required. This data visualization can be more effectively and efficiently customized and optimized through specialized services or solutions based on the user's business, industry, technology, and problem. Therefore, Mellerikat focuses on operating and optimizing AI models and delivering standardized inference results to users, who then use the provided inference results to conduct analysis/visualization suitable for their needs. To this end, Mellerikat regulates AI Solutions to provide standardized forms of summarized/detailed information for inference results.