Object Detection
Feature Description
Object detection utilizes deep learning neural network models to automatically determine and identify object categories and locations in images. It can detect multiple object categories simultaneously and output results in XML format, which records object categories, locations, and confidence scores. This feature demonstrates fast detection speed and high accuracy.
Parameter Description
| Parameter Name | Default Value | Description | Type |
|---|---|---|---|
| File Type | Dataset | Specify file type. Available options: dataset or folder. | Boolean |
| Model File | Specify model file (*.sdm) for image object detection | String | |
| Probability Threshold | 0.5 | The system calculates probability scores reflecting object feature compliance. Only objects with scores above this threshold will be saved. Default: 0.5. | Double |
| Deduplicate Threshold | 0.3 | Specifies threshold for duplicate removal | Double |
| Batch Size | 1 | Specifies number of images processed per detection batch | Integer |
| Processor Type | GPU | Specifies processor type | String |
| GPU ID | 0 | Input GPU ID(s) for multi-card inference. Multiple IDs should be comma-separated (e.g., "0,1,2" will be converted to "012") | String |
| Additional Parameters (Optional) |
false | Check this box to enable deduplicate threshold configuration | Boolean |
| Result Datasource | Specifies datasource containing result dataset | Datasource | |
| Result Dataset | ObjectDetectionResult | Specifies name for output dataset | String |
Output
| Parameter Name | Description | Type |
|---|---|---|
| Result Dataset | Output result dataset | DatasetVector |