Detections Stabilizer

Class: StabilizeTrackedDetectionsBlockV1

Source: inference.core.workflows.core_steps.transformations.stabilize_detections.v1.StabilizeTrackedDetectionsBlockV1

This block stores last known position for each bounding box If box disappears then this block will bring it back so short gaps are filled with last known box position The block requires detections to be tracked (i.e. each object must have unique tracker_id assigned, which persists between frames) WARNING: this block will produce many short-lived bounding boxes for unstable trackers!

Type identifier

Use the following identifier in step "type" field: roboflow_core/stabilize_detections@v1to add the block as as step in your workflow.

Properties

Name	Type	Description	Refs
name	str	Enter a unique identifier for this step..	❌
smoothing_window_size	int	Predicted movement of detection will be smoothed based on historical measurements of velocity, this parameter controls number of historical measurements taken under account when calculating smoothed velocity. Detections will be removed from generating smoothed predictions if they had been missing for longer than this number of frames..	✅
bbox_smoothing_coefficient	float	Bounding box smoothing coefficient applied when given tracker_id is present on current frame. This parameter must be initialized with value between 0 and 1.	✅

The Refs column marks possibility to parametrise the property with dynamic values available in workflow runtime. See Bindings for more info.

Available Connections

Compatible Blocks

Check what blocks you can connect to Detections Stabilizer in version v1.

• inputs: Byte Tracker, Polygon Zone Visualization, Time in Zone, Distance Measurement, Identify Outliers, Dot Visualization, Morphological Transformation, Blur Visualization, Perspective Correction, Corner Visualization, Pixel Color Count, PTZ Tracking (ONVIF).md), Grid Visualization, Image Threshold, Halo Visualization, Detections Combine, Detection Offset, EasyOCR, Byte Tracker, Line Counter Visualization, VLM as Detector, Stability AI Outpainting, Google Vision OCR, Identify Changes, Camera Focus, SIFT, Instance Segmentation Model, Clip Comparison, Image Slicer, Image Convert Grayscale, Keypoint Visualization, Template Matching, OCR Model, Detections Stabilizer, Bounding Box Visualization, Instance Segmentation Model, Line Counter, Reference Path Visualization, Dynamic Crop, Mask Visualization, Image Preprocessing, Background Color Visualization, Camera Calibration, Depth Estimation, Image Slicer, QR Code Generator, Detections Stitch, SIFT Comparison, Trace Visualization, Time in Zone, Velocity, Object Detection Model, Detections Transformation, Contrast Equalization, Byte Tracker, Overlap Filter, Crop Visualization, Stability AI Image Generation, Moondream2, SIFT Comparison, Object Detection Model, Model Comparison Visualization, Pixelate Visualization, Dynamic Zone, Line Counter, Time in Zone, Relative Static Crop, Image Contours, Polygon Visualization, Path Deviation, Triangle Visualization, Detections Filter, YOLO-World Model, Classification Label Visualization, Detections Classes Replacement, Bounding Rectangle, Circle Visualization, Image Blur, Label Visualization, VLM as Detector, Absolute Static Crop, Stability AI Inpainting, Icon Visualization, Ellipse Visualization, Color Visualization, Detections Merge, Path Deviation, Detections Consensus, Stitch Images, Segment Anything 2 Model
• outputs: Byte Tracker, Distance Measurement, Time in Zone, Dot Visualization, Detections Stitch, Size Measurement, Time in Zone, Blur Visualization, Perspective Correction, Velocity, Trace Visualization, Corner Visualization, Detections Transformation, Byte Tracker, Overlap Filter, PTZ Tracking (ONVIF).md), Florence-2 Model, Crop Visualization, Florence-2 Model, Halo Visualization, Roboflow Custom Metadata, Detections Combine, Detection Offset, Model Comparison Visualization, Pixelate Visualization, Dynamic Zone, Byte Tracker, Model Monitoring Inference Aggregator, Line Counter, Time in Zone, Stitch OCR Detections, Polygon Visualization, Path Deviation, Triangle Visualization, Roboflow Dataset Upload, Detections Filter, Detections Classes Replacement, Bounding Rectangle, Circle Visualization, Detections Stabilizer, Bounding Box Visualization, Line Counter, Label Visualization, Stability AI Inpainting, Icon Visualization, Ellipse Visualization, Dynamic Crop, Color Visualization, Roboflow Dataset Upload, Mask Visualization, Detections Merge, Segment Anything 2 Model, Path Deviation, Detections Consensus, Background Color Visualization

Input and Output Bindings

The available connections depend on its binding kinds. Check what binding kinds Detections Stabilizer in version v1 has.

Bindings

• input

  • image (image): not available.
  • detections (Union[object_detection_prediction, instance_segmentation_prediction]): Tracked detections.
  • smoothing_window_size (integer): Predicted movement of detection will be smoothed based on historical measurements of velocity, this parameter controls number of historical measurements taken under account when calculating smoothed velocity. Detections will be removed from generating smoothed predictions if they had been missing for longer than this number of frames..
  • bbox_smoothing_coefficient (float_zero_to_one): Bounding box smoothing coefficient applied when given tracker_id is present on current frame. This parameter must be initialized with value between 0 and 1.

• output

  • tracked_detections (Union[object_detection_prediction, instance_segmentation_prediction]): Prediction with detected bounding boxes in form of sv.Detections(...) object if object_detection_prediction or Prediction with detected bounding boxes and segmentation masks in form of sv.Detections(...) object if instance_segmentation_prediction.

Example JSON definition of step Detections Stabilizer in version v1

{
    "name": "<your_step_name_here>",
    "type": "roboflow_core/stabilize_detections@v1",
    "image": "<block_does_not_provide_example>",
    "detections": "$steps.object_detection_model.predictions",
    "smoothing_window_size": 5,
    "bbox_smoothing_coefficient": 0.2
}