Deep Learning Result Parser
Function
This Step is used to parse the inference results of a multi-model package or a multi-class defect segmentation model package output by the "Deep Learning Model Package Inference" Step. After parsing, this Step splits the output results by model package type or defect class, facilitating independent processing, statistics, and viewing by subsequent Steps.
Usage Scenario
This Step follows the "Deep Learning Model Package Inference" Step and is used to parse and view the inference results of a multi-model package or a multi-class defect segmentation model package.
|
This Step can only parse the inference results of one defect class/model package type at a time. To parse the inference results of multiple defect classes or model packages simultaneously, please add and connect a separate instance of this Step for each defect class/model package output by the "Deep Learning Model Package Inference" Step to avoid missing any parsing results. |
Input and Output
Input Description
| Input item | Description |
|---|---|
Deep Learning Inference Result |
Deep learning inference result. |
Output Description
The output ports are automatically generated based on the defect classes or model package types in the input data. No manual configuration is required. Please ensure the input data is correct to avoid unexpected output results.
Usage Examples
Multi-Class Defect Segmentation Result Parsing
The figure below shows how to parse the multi-class defect segmentation results output by the "Deep Learning Model Package Inference" Step. Taking a model that simultaneously detects scratches, bubbles, and dents as an example, each defect class is connected to a separate "Deep Learning Result Parser" Step for independent parsing.
Multi-Model Package Inference Result Parsing
The figure below shows how to parse the multi-model package inference results output by the "Deep Learning Model Package Inference" Step. Taking a multi-model package with text detection and text recognition in series as an example, each model package type is connected to a separate "Deep Learning Result Parser" Step for independent parsing.
