SAM 3

Class: SegmentAnything3BlockV1

Source: inference.core.workflows.core_steps.models.foundation.segment_anything3.v1.SegmentAnything3BlockV1

Run Segment Anything 3, a zero-shot instance segmentation model, on an image.

You can pass in boxes/predictions from other models as prompts, or use a text prompt for open-vocabulary segmentation. If you pass in box detections from another model, the class names of the boxes will be forwarded to the predicted masks.

Type identifier

Use the following identifier in step "type" field: roboflow_core/sam3@v1to add the block as as step in your workflow.

Properties

Name	Type	Description	Refs
name	str	Enter a unique identifier for this step..	❌
model_id	str	model version. You only need to change this for fine tuned sam3 models..	✅
class_names	Optional[List[str], str]	List of classes to recognise.	✅
threshold	float	Threshold for predicted mask scores.	✅

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 SAM 3 in version v1.

• inputs: LMM, Background Color Visualization, Stitch Images, Image Slicer, Size Measurement, Dynamic Zone, Model Monitoring Inference Aggregator, Corner Visualization, Camera Calibration, Mask Visualization, Object Detection Model, Local File Sink, Model Comparison Visualization, Keypoint Detection Model, Email Notification, Pixelate Visualization, Anthropic Claude, Relative Static Crop, Google Gemini, Florence-2 Model, Multi-Label Classification Model, Ellipse Visualization, Triangle Visualization, Camera Focus, OCR Model, QR Code Generator, Label Visualization, Roboflow Custom Metadata, Florence-2 Model, LMM For Classification, Identify Changes, Blur Visualization, Single-Label Classification Model, Dot Visualization, Stability AI Image Generation, Perspective Correction, Google Vision OCR, Llama 3.2 Vision, EasyOCR, Absolute Static Crop, Slack Notification, Morphological Transformation, Cosine Similarity, Image Blur, Image Threshold, Clip Comparison, Stitch OCR Detections, Depth Estimation, Stability AI Outpainting, Halo Visualization, Stability AI Inpainting, Polygon Visualization, OpenAI, Grid Visualization, Roboflow Dataset Upload, CogVLM, Classification Label Visualization, Email Notification, VLM as Detector, Instance Segmentation Model, Bounding Box Visualization, Image Convert Grayscale, Polygon Zone Visualization, OpenAI, Clip Comparison, Keypoint Detection Model, Crop Visualization, Object Detection Model, Image Slicer, Multi-Label Classification Model, Gaze Detection, Icon Visualization, Color Visualization, Roboflow Dataset Upload, Dimension Collapse, Keypoint Visualization, Contrast Equalization, Buffer, Image Contours, Instance Segmentation Model, Circle Visualization, OpenAI, VLM as Classifier, CSV Formatter, Reference Path Visualization, Twilio SMS Notification, Dynamic Crop, Webhook Sink, Single-Label Classification Model, SIFT, Line Counter Visualization, Image Preprocessing, Trace Visualization, SIFT Comparison
• outputs: Polygon Visualization, Roboflow Dataset Upload, Background Color Visualization, Size Measurement, Dynamic Zone, Path Deviation, Detections Transformation, Corner Visualization, Model Monitoring Inference Aggregator, Detections Classes Replacement, Distance Measurement, Trace Visualization, Mask Visualization, Line Counter, Bounding Box Visualization, Model Comparison Visualization, Byte Tracker, Pixelate Visualization, Time in Zone, Florence-2 Model, Detections Stitch, Crop Visualization, Triangle Visualization, Segment Anything 2 Model, Ellipse Visualization, Detections Merge, Icon Visualization, Byte Tracker, Color Visualization, Roboflow Dataset Upload, Path Deviation, Label Visualization, Time in Zone, Roboflow Custom Metadata, Florence-2 Model, Blur Visualization, Dot Visualization, Overlap Filter, Detections Filter, Circle Visualization, Detection Offset, Perspective Correction, Bounding Rectangle, Line Counter, Detections Combine, Detections Stabilizer, Time in Zone, Velocity, Detections Consensus, PTZ Tracking (ONVIF).md), Byte Tracker, Halo Visualization, Stability AI Inpainting, Dynamic Crop

Input and Output Bindings

The available connections depend on its binding kinds. Check what binding kinds SAM 3 in version v1 has.

Bindings

• input

  • images (image): The image to infer on..
  • model_id (roboflow_model_id): model version. You only need to change this for fine tuned sam3 models..
  • class_names (Union[string, list_of_values]): List of classes to recognise.
  • threshold (float): Threshold for predicted mask scores.

• output

  • predictions (instance_segmentation_prediction): Prediction with detected bounding boxes and segmentation masks in form of sv.Detections(...) object.

Example JSON definition of step SAM 3 in version v1

{
    "name": "<your_step_name_here>",
    "type": "roboflow_core/sam3@v1",
    "images": "$inputs.image",
    "model_id": "sam3/sam3_final",
    "class_names": [
        "car",
        "person"
    ],
    "threshold": 0.3
}