NPU Components
The following nodes can only be used in combination with a VIU2.
NPU Yolo Classification
This component allows you to classify images using AI methods. Configure the basic settings and use the Get node of this component in combination with the "Classify Image (NPU)" node.
Flow
Parameter set
Attribute | Type | Description |
|---|---|---|
Classes File Path | FileStoreItem | Specifies the path to the JSON file containing the possible classes. |
Return Preprocessed Image | Boolean | Determines whether the pre-processed image should be returned, as this incurs runtime costs. |
Model File Path | FileStoreItem | Specifies the path to the quantized TFLite model to be used for inference. |
Number Of Threads | Int32 | Specifies the number of threads to use for inference. |
Confidence Threshold | Single | Sets the minimum confidence level an object must have to be considered recognized. The value must be between 0 and 1. |
Maximal Number Of Detections | Int32 | Sets the maximum number of returned detections. |
NPU Yolo Detection
This component allows you to detect objects using AI methods. Configure the basic settings and use the Get node of this component in combination with the "Detect Objects (NPU)" node.
Flow
Parameter set
Attribute | Type | Description |
|---|---|---|
Classes File Path | FileStoreItem | Specifies the path to the JSON file containing the possible classes. |
Return Preprocessed Image | Boolean | Determines whether the pre-processed image should be returned, as this incurs runtime costs. |
Model File Path | FileStoreItem | Specifies the path to the quantized TFLite model to be used for inference. |
Number Of Threads | Int32 | Specifies the number of threads to use for inference. |
Oriented Boxes | Boolean | Determines whether object recognition is performed normally or in an oriented manner. |
Confidence Threshold | Single | Sets the minimum confidence level an object must reach to be considered recognized. The value must be between 0 and 1. |
Iou Threshold | Single | Sets the maximum IoU value between two detected objects. If the threshold is exceeded, the objects are counted as one. The value must be between 0 and 1. |
Maximal Number Of Detections | Int32 | Defines the maximum number of returned detections. |
Draw Object Name | Boolean | Determines whether the object name should be displayed in the processed image. |
Draw Object Score | Boolean | Determines whether the confidence score should be displayed in the processed image. |
Bounding Box Thickness | Int32 | Defines the thickness of the bounding boxes in the processed image. The value must be greater than 0. |
Object Name Thickness | Int32 | Defines the thickness of the bounding box label. The value must be greater than 0. |
Object Name Font Size | Single | Defines the text size of the label within the bounding box. |
NPU Yolo Segmentation
This component allows you to perform instance segmentation using AI methods. Configure the basic settings and use the Get node of this component in combination with the "Segment Instances (NPU)" node.
Flow
Parameter set
Attribute | Type | Description |
|---|---|---|
Bilinear Interpolation | Boolean | Determines whether bilinear interpolation should be performed in post-processing. This increases runtime but results in significantly smoother masks. |
Classes File Path | FileStoreItem | Specifies the path to the JSON file containing the possible classes. |
Return Preprocessed Image | Boolean | Determines whether the pre-processed image should be returned. This consumes processing time. |
Model File Path | FileStoreItem | Specifies the path to the ONNX model to be used for inference. |
Number Of Threads | Int32 | Specifies the number of threads to be used for inference. |
Confidence Threshold | Single | Sets the minimum confidence level an object must have to be considered recognized. The value must be between 0 and 1. |
Mask Threshold | Single | Sets the threshold that determines whether a pixel is part of the object mask or not. The value must be between 0 and 1. |
Iou Threshold | Single | Defines the maximum IoU value between two detected objects. If the threshold is exceeded, the objects are counted as one. The value must be between 0 and 1. |
Maximal Number Of Detections | Int32 | Defines the maximum number of returned detections. |
Draw Object Name | Boolean | Determines whether the object name should be displayed in the processed image. |
Draw Object Score | Boolean | Determines whether the confidence score should be displayed in the processed image. |
Bounding Box Thickness | Int32 | Defines the thickness of the bounding boxes in the processed image. The value must be greater than 0. |
Object Name Thickness | Int32 | Defines the thickness of the bounding box label. The value must be greater than 0. |
Object Name Font Size | Single | Defines the text size of the label within the bounding box. |