Multi-Model Package
Function Introduction
Use a multi-model package for inference on input images. A multi-model package can integrate multiple sub-models (such as image classification, object detection, defect segmentation, and OCR) and execute them with predefined serial, parallel, or hybrid logic. It outputs both sub-model results and overall verification status.
This is suitable for complex quality-inspection scenarios, reduces number of projects, avoids repeated model configuration, and improves deployment/maintenance efficiency.
Input and Output
After importing this model package in Deep Learning Model Package Inference, the following ports are available.
Input
| Input Port | Data Type | Description |
|---|---|---|
Image |
Image/Color |
Input image used for model-package inference. |
Output
| Output Port | Data Type | Description |
|---|---|---|
Overall Verification Result |
Bool |
Overall verification status of combined model results. |
Image Classification |
DLResult/Classification |
Model-package inference result. |
Fast Positioning |
DLResult/FastLocating |
Model-package inference result. |
Text Detection |
DLResult/TextDetection |
Model-package inference result. |
Unsupervised Segmentation |
DLResult/UnsupervisedSegmentation |
Model-package inference result. |
Defect Segmentation |
DLResult/DefectSegmentation |
Model-package inference result. |
Object Detection |
DLResult/ObjectDetection |
Model-package inference result. |
Instance Segmentation |
DLResult/InstanceSegmentation |
Model-package inference result. |
Text Recognition |
DLResult/TextRecognition |
Model-package inference result. |
Parameter Description
When a multi-model package is imported, configure the following parameters.
Model Package Settings
| Parameter | Description |
|---|---|
Model Package Management Tool |
Opens model package management tool to import |
Model Package Name |
Selects imported model package for current step. |
Release Previous Package After Switching |
Controls whether resources of previous package are released immediately after switching. Default value: Enabled. |
Model Package Type |
Auto-filled after selecting model package name. |
GPU ID |
Selects GPU device ID used for package inference. |
Pre-Processing
| Parameter | Description |
|---|---|
ROI File |
Sets or edits ROI of input image. [CAUTION] ==== Before inference, verify ROI here is consistent with ROI configured in Mech-DLK. Inconsistent ROI may affect recognition performance. ==== [TIP] ==== To restore default ROI, clear ROI name under Open Editor. ==== |
Post-Processing
| Parameter | Description |
|---|---|
Inference Configuration |
Configures inference parameters for multi-model package. Click Open Editor to open configuration window. |
Category Display Mode |
Chooses category display by name or by index in output. |
Visualization Settings
| Parameter | Description |
|---|---|
Visualize Detection Result |
Displays detection result on image when enabled. Default value: Disabled. |
Visualization Mode |
Defines visualization style in output. Default value: Show Each Instance Value list: Show Each Instance, Show Instances by Class, Show Instance Centers. |
Use Custom Font Size |
Enables custom text size in visualization output. Default value: Disabled. |
Font Size (0-10) |
Sets text size in visualization output. Default value: 3.0 |
