Instance Segmentation Model¶
v4¶
Class: RoboflowInstanceSegmentationModelBlockV4 (there are multiple versions of this block)
Warning: This block has multiple versions. Please refer to the specific version for details. You can learn more about how versions work here: Versioning
Run inference on an instance segmentation model hosted on or uploaded to Roboflow.
You can query any model that is private to your account, or any public model available on Roboflow Universe.
You will need to set your Roboflow API key in your Inference environment to use this block. To learn more about setting your Roboflow API key, refer to the Inference documentation.
This version of block introduces breaking change in behaviour of mask construction - it uses
rle format instead polygon making it possible to retrieve
shapes of any kind from remote server.
Type identifier¶
Use the following identifier in step "type" field: roboflow_core/roboflow_instance_segmentation_model@v4to 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 |
Roboflow model identifier.. | ✅ |
confidence_mode |
str |
How confidence thresholds are determined.. | ✅ |
custom_confidence |
float |
Custom confidence threshold for predictions.. | ✅ |
class_filter |
List[str] |
List of accepted classes. Classes must exist in the model's training set.. | ✅ |
iou_threshold |
float |
Minimum overlap threshold between boxes to combine them into a single detection, used in NMS. Learn more.. | ✅ |
max_detections |
int |
Maximum number of detections to return.. | ✅ |
class_agnostic_nms |
bool |
Boolean flag to specify if NMS is to be used in class-agnostic mode.. | ✅ |
max_candidates |
int |
Maximum number of candidates as NMS input to be taken into account.. | ✅ |
mask_decode_mode |
str |
Parameter of mask decoding in prediction post-processing.. | ✅ |
tradeoff_factor |
float |
Post-processing parameter to dictate tradeoff between fast and accurate.. | ✅ |
disable_active_learning |
bool |
Boolean flag to disable project-level active learning for this block.. | ✅ |
active_learning_target_dataset |
str |
Target dataset for active learning, if enabled.. | ✅ |
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 Instance Segmentation Model in version v4.
- inputs:
Template Matching,Morphological Transformation,Classification Label Visualization,Crop Visualization,Stability AI Outpainting,Blur Visualization,Reference Path Visualization,OpenAI,Anthropic Claude,Camera Focus,Instance Segmentation Model,Size Measurement,Model Comparison Visualization,Florence-2 Model,Trace Visualization,JSON Parser,Label Visualization,Image Convert Grayscale,Florence-2 Model,Text Display,Qwen-VL,Llama 3.2 Vision,Image Blur,Keypoint Detection Model,Absolute Static Crop,CSV Formatter,Keypoint Detection Model,LMM,Qwen 3.5 API,Qwen 3.6 API,Camera Focus,Line Counter,VLM As Detector,Multi-Label Classification Model,Clip Comparison,Google Gemma API,Contrast Enhancement,Halo Visualization,Color Visualization,Morphological Transformation,MoonshotAI Kimi,Stitch OCR Detections,Event Writer,Buffer,Stability AI Inpainting,Roboflow Asset Library Attributes,Microsoft SQL Server Sink,OpenAI,Roboflow Vision Events,Identify Outliers,CogVLM,Detections Consensus,Object Detection Model,OPC UA Writer Sink,Semantic Segmentation Model,Dynamic Crop,Bounding Box Visualization,Qwen3.5-VL,Clip Comparison,OpenAI,SIFT Comparison,OCR Model,Single-Label Classification Model,Slack Notification,OpenRouter,Detection Event Log,SIFT Comparison,Pixelate Visualization,Google Vision OCR,Google Gemma,Dynamic Zone,Halo Visualization,Stitch OCR Detections,GLM-OCR,Image Threshold,Stitch Images,Twilio SMS/MMS Notification,Icon Visualization,VLM As Classifier,MoonshotAI Kimi,Single-Label Classification Model,Google Gemini,Single-Label Classification Model,Webhook Sink,Instance Segmentation Model,QR Code Generator,MQTT Writer,Ellipse Visualization,Object Detection Model,Anthropic Claude,Keypoint Detection Model,Dot Visualization,Perspective Correction,Instance Segmentation Model,Roboflow Dataset Upload,PLC ModbusTCP,SIFT,Google Gemini,EasyOCR,Dimension Collapse,Local File Sink,Triangle Visualization,Contrast Equalization,Polygon Visualization,OpenAI,Heatmap Visualization,Detections List Roll-Up,Google Gemini,PLC EthernetIP,LMM For Classification,VLM As Detector,Llama 3.2 Vision,Multi-Label Classification Model,Polygon Visualization,Email Notification,Identify Changes,Mask Visualization,Anthropic Claude,Image Stack,Distance Measurement,PTZ Tracking (ONVIF),Keypoint Visualization,Background Subtraction,Multi-Label Classification Model,Twilio SMS Notification,Email Notification,Semantic Segmentation Model,Image Slicer,Image Contours,Line Counter Visualization,Image Preprocessing,VLM As Classifier,Depth Estimation,Pixel Color Count,Motion Detection,Current Time,Corner Visualization,Polygon Zone Visualization,Camera Calibration,Roboflow Dataset Upload,Grid Visualization,Stability AI Image Generation,S3 Sink,Circle Visualization,Image Slicer,Roboflow Custom Metadata,Relative Static Crop,Instance Segmentation Model,Model Monitoring Inference Aggregator,OpenAI-Compatible LLM,Object Detection Model,Background Color Visualization,Line Counter - outputs:
Halo Visualization,Overlap Analysis,GLM-OCR,SAM 3 Interactive,Crop Visualization,Icon Visualization,Detections Transformation,Blur Visualization,ByteTrack Tracker,Detections Classes Replacement,Single-Label Classification Model,Byte Tracker,Single-Label Classification Model,Webhook Sink,Instance Segmentation Model,Track Class Lock,Size Measurement,Mask Edge Snap,Instance Segmentation Model,Model Comparison Visualization,Path Deviation,Florence-2 Model,Trace Visualization,Ellipse Visualization,Object Detection Model,SmolVLM2,Keypoint Detection Model,BoT-SORT Tracker,Dot Visualization,Perspective Correction,Label Visualization,Instance Segmentation Model,Florence-2 Model,Per-Class Confidence Filter,Qwen-VL,Roboflow Dataset Upload,Detections Stabilizer,Keypoint Detection Model,Detections Merge,Velocity,Keypoint Detection Model,OC-SORT Tracker,Qwen2.5-VL,SAM 3,Triangle Visualization,Camera Focus,Time in Zone,SORT Tracker,Polygon Visualization,Line Counter,SAM2 Video Tracker,Qwen3-VL,Heatmap Visualization,Multi-Label Classification Model,Detections Stitch,Detections List Roll-Up,Halo Visualization,Color Visualization,Event Writer,Multi-Label Classification Model,Polygon Visualization,Mask Visualization,Detections Filter,Distance Measurement,Stability AI Inpainting,Bounding Rectangle,PTZ Tracking (ONVIF),Time in Zone,Overlap Filter,Roboflow Vision Events,Multi-Label Classification Model,Mask Area Measurement,Semantic Segmentation Model,Detection Offset,Detections Consensus,Object Detection Model,Byte Tracker,SAM 3,Semantic Segmentation Model,Dynamic Crop,Path Deviation,Byte Tracker,Bounding Box Visualization,Detections Combine,Qwen3.5-VL,Qwen3.5,Corner Visualization,Roboflow Dataset Upload,Moondream2,Segment Anything 2 Model,SAM 3,Circle Visualization,Time in Zone,Single-Label Classification Model,Roboflow Custom Metadata,Instance Segmentation Model,Model Monitoring Inference Aggregator,Object Detection Model,Detection Event Log,Pixelate Visualization,Background Color Visualization,Line Counter,SAM3 Video Tracker,Dynamic Zone
Input and Output Bindings¶
The available connections depend on its binding kinds. Check what binding kinds
Instance Segmentation Model in version v4 has.
Bindings
-
input
images(image): The image to infer on..model_id(roboflow_model_id): Roboflow model identifier..confidence_mode(string): How confidence thresholds are determined..custom_confidence(float_zero_to_one): Custom confidence threshold for predictions..class_filter(list_of_values): List of accepted classes. Classes must exist in the model's training set..iou_threshold(float_zero_to_one): Minimum overlap threshold between boxes to combine them into a single detection, used in NMS. Learn more..max_detections(integer): Maximum number of detections to return..class_agnostic_nms(boolean): Boolean flag to specify if NMS is to be used in class-agnostic mode..max_candidates(integer): Maximum number of candidates as NMS input to be taken into account..mask_decode_mode(string): Parameter of mask decoding in prediction post-processing..tradeoff_factor(float_zero_to_one): Post-processing parameter to dictate tradeoff between fast and accurate..disable_active_learning(boolean): Boolean flag to disable project-level active learning for this block..active_learning_target_dataset(roboflow_project): Target dataset for active learning, if enabled..
-
output
inference_id(inference_id): Inference identifier.predictions(Union[rle_instance_segmentation_prediction,instance_segmentation_prediction]): Prediction with detected bounding boxes and RLE-encoded segmentation masks in form of sv.Detections(...) object ifrle_instance_segmentation_predictionor Prediction with detected bounding boxes and segmentation masks in form of sv.Detections(...) object ifinstance_segmentation_prediction.model_id(roboflow_model_id): Roboflow model id.
Example JSON definition of step Instance Segmentation Model in version v4
{
"name": "<your_step_name_here>",
"type": "roboflow_core/roboflow_instance_segmentation_model@v4",
"images": "$inputs.image",
"model_id": "my_project/3",
"confidence_mode": "<block_does_not_provide_example>",
"custom_confidence": 0.3,
"class_filter": [
"a",
"b",
"c"
],
"iou_threshold": 0.4,
"max_detections": 300,
"class_agnostic_nms": true,
"max_candidates": 3000,
"mask_decode_mode": "accurate",
"tradeoff_factor": 0.3,
"disable_active_learning": true,
"active_learning_target_dataset": "my_project"
}
v3¶
Class: RoboflowInstanceSegmentationModelBlockV3 (there are multiple versions of this block)
Warning: This block has multiple versions. Please refer to the specific version for details. You can learn more about how versions work here: Versioning
Run inference on an instance segmentation model hosted on or uploaded to Roboflow.
You can query any model that is private to your account, or any public model available on Roboflow Universe.
You will need to set your Roboflow API key in your Inference environment to use this block. To learn more about setting your Roboflow API key, refer to the Inference documentation.
Type identifier¶
Use the following identifier in step "type" field: roboflow_core/roboflow_instance_segmentation_model@v3to 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 |
Roboflow model identifier.. | ✅ |
confidence_mode |
str |
How confidence thresholds are determined.. | ✅ |
custom_confidence |
float |
Custom confidence threshold for predictions.. | ✅ |
class_filter |
List[str] |
List of accepted classes. Classes must exist in the model's training set.. | ✅ |
iou_threshold |
float |
Minimum overlap threshold between boxes to combine them into a single detection, used in NMS. Learn more.. | ✅ |
max_detections |
int |
Maximum number of detections to return.. | ✅ |
class_agnostic_nms |
bool |
Boolean flag to specify if NMS is to be used in class-agnostic mode.. | ✅ |
max_candidates |
int |
Maximum number of candidates as NMS input to be taken into account.. | ✅ |
mask_decode_mode |
str |
Parameter of mask decoding in prediction post-processing.. | ✅ |
tradeoff_factor |
float |
Post-processing parameter to dictate tradeoff between fast and accurate.. | ✅ |
disable_active_learning |
bool |
Boolean flag to disable project-level active learning for this block.. | ✅ |
active_learning_target_dataset |
str |
Target dataset for active learning, if enabled.. | ✅ |
enforce_dense_masks_in_inference_models |
bool |
Boolean flag to enforce dense masks when inference models backend is in use (irrelevant in other cases). Dense masks are faster to process, but require more memory. Users can't tweak this flag when running on Roboflow serverless platform.. | ✅ |
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 Instance Segmentation Model in version v3.
- inputs:
Template Matching,Morphological Transformation,Classification Label Visualization,Crop Visualization,Stability AI Outpainting,Blur Visualization,Reference Path Visualization,OpenAI,Anthropic Claude,Camera Focus,Instance Segmentation Model,Size Measurement,Model Comparison Visualization,Florence-2 Model,Trace Visualization,JSON Parser,Label Visualization,Image Convert Grayscale,Florence-2 Model,Text Display,Qwen-VL,Llama 3.2 Vision,Image Blur,Keypoint Detection Model,Absolute Static Crop,CSV Formatter,Keypoint Detection Model,LMM,Qwen 3.5 API,Qwen 3.6 API,Camera Focus,Line Counter,VLM As Detector,Multi-Label Classification Model,Clip Comparison,Google Gemma API,Contrast Enhancement,Halo Visualization,Color Visualization,Morphological Transformation,MoonshotAI Kimi,Stitch OCR Detections,Event Writer,Buffer,Stability AI Inpainting,Roboflow Asset Library Attributes,Microsoft SQL Server Sink,OpenAI,Roboflow Vision Events,Identify Outliers,CogVLM,Detections Consensus,Object Detection Model,OPC UA Writer Sink,Semantic Segmentation Model,Dynamic Crop,Bounding Box Visualization,Qwen3.5-VL,Clip Comparison,OpenAI,SIFT Comparison,OCR Model,Single-Label Classification Model,Slack Notification,OpenRouter,Detection Event Log,SIFT Comparison,Pixelate Visualization,Google Vision OCR,Google Gemma,Dynamic Zone,Halo Visualization,Stitch OCR Detections,GLM-OCR,Image Threshold,Stitch Images,Twilio SMS/MMS Notification,Icon Visualization,VLM As Classifier,MoonshotAI Kimi,Single-Label Classification Model,Google Gemini,Single-Label Classification Model,Webhook Sink,Instance Segmentation Model,QR Code Generator,MQTT Writer,Ellipse Visualization,Object Detection Model,Anthropic Claude,Keypoint Detection Model,Dot Visualization,Perspective Correction,Instance Segmentation Model,Roboflow Dataset Upload,PLC ModbusTCP,SIFT,Google Gemini,EasyOCR,Dimension Collapse,Local File Sink,Triangle Visualization,Contrast Equalization,Polygon Visualization,OpenAI,Heatmap Visualization,Detections List Roll-Up,Google Gemini,PLC EthernetIP,LMM For Classification,VLM As Detector,Llama 3.2 Vision,Multi-Label Classification Model,Polygon Visualization,Email Notification,Identify Changes,Mask Visualization,Anthropic Claude,Image Stack,Distance Measurement,PTZ Tracking (ONVIF),Keypoint Visualization,Background Subtraction,Multi-Label Classification Model,Twilio SMS Notification,Email Notification,Semantic Segmentation Model,Image Slicer,Image Contours,Line Counter Visualization,Image Preprocessing,VLM As Classifier,Depth Estimation,Pixel Color Count,Motion Detection,Current Time,Corner Visualization,Polygon Zone Visualization,Camera Calibration,Roboflow Dataset Upload,Grid Visualization,Stability AI Image Generation,S3 Sink,Circle Visualization,Image Slicer,Roboflow Custom Metadata,Relative Static Crop,Instance Segmentation Model,Model Monitoring Inference Aggregator,OpenAI-Compatible LLM,Object Detection Model,Background Color Visualization,Line Counter - outputs:
Halo Visualization,Overlap Analysis,GLM-OCR,SAM 3 Interactive,Crop Visualization,Icon Visualization,Detections Transformation,Blur Visualization,ByteTrack Tracker,Detections Classes Replacement,Single-Label Classification Model,Byte Tracker,Single-Label Classification Model,Webhook Sink,Instance Segmentation Model,Track Class Lock,Size Measurement,Mask Edge Snap,Instance Segmentation Model,Model Comparison Visualization,Path Deviation,Florence-2 Model,Trace Visualization,Ellipse Visualization,Object Detection Model,SmolVLM2,Keypoint Detection Model,BoT-SORT Tracker,Dot Visualization,Perspective Correction,Label Visualization,Instance Segmentation Model,Florence-2 Model,Per-Class Confidence Filter,Qwen-VL,Roboflow Dataset Upload,Detections Stabilizer,Keypoint Detection Model,Detections Merge,Velocity,Keypoint Detection Model,OC-SORT Tracker,Qwen2.5-VL,SAM 3,Triangle Visualization,Camera Focus,Time in Zone,Line Counter,SORT Tracker,SAM2 Video Tracker,Polygon Visualization,Qwen3-VL,Heatmap Visualization,Multi-Label Classification Model,Detections Stitch,Detections List Roll-Up,Halo Visualization,Color Visualization,Event Writer,Multi-Label Classification Model,Polygon Visualization,Mask Visualization,Detections Filter,Distance Measurement,Stability AI Inpainting,Bounding Rectangle,PTZ Tracking (ONVIF),Time in Zone,Overlap Filter,Roboflow Vision Events,Multi-Label Classification Model,Mask Area Measurement,Semantic Segmentation Model,Detection Offset,Detections Consensus,Object Detection Model,Byte Tracker,SAM 3,Semantic Segmentation Model,Dynamic Crop,Path Deviation,Byte Tracker,Bounding Box Visualization,Detections Combine,Qwen3.5-VL,Qwen3.5,Roboflow Dataset Upload,Corner Visualization,Moondream2,Segment Anything 2 Model,SAM 3,Circle Visualization,Time in Zone,Single-Label Classification Model,Roboflow Custom Metadata,Instance Segmentation Model,Model Monitoring Inference Aggregator,Object Detection Model,Detection Event Log,Pixelate Visualization,Background Color Visualization,Line Counter,SAM3 Video Tracker,Dynamic Zone
Input and Output Bindings¶
The available connections depend on its binding kinds. Check what binding kinds
Instance Segmentation Model in version v3 has.
Bindings
-
input
images(image): The image to infer on..model_id(roboflow_model_id): Roboflow model identifier..confidence_mode(string): How confidence thresholds are determined..custom_confidence(float_zero_to_one): Custom confidence threshold for predictions..class_filter(list_of_values): List of accepted classes. Classes must exist in the model's training set..iou_threshold(float_zero_to_one): Minimum overlap threshold between boxes to combine them into a single detection, used in NMS. Learn more..max_detections(integer): Maximum number of detections to return..class_agnostic_nms(boolean): Boolean flag to specify if NMS is to be used in class-agnostic mode..max_candidates(integer): Maximum number of candidates as NMS input to be taken into account..mask_decode_mode(string): Parameter of mask decoding in prediction post-processing..tradeoff_factor(float_zero_to_one): Post-processing parameter to dictate tradeoff between fast and accurate..disable_active_learning(boolean): Boolean flag to disable project-level active learning for this block..active_learning_target_dataset(roboflow_project): Target dataset for active learning, if enabled..enforce_dense_masks_in_inference_models(boolean): Boolean flag to enforce dense masks when inference models backend is in use (irrelevant in other cases). Dense masks are faster to process, but require more memory. Users can't tweak this flag when running on Roboflow serverless platform..
-
output
inference_id(inference_id): Inference identifier.predictions(instance_segmentation_prediction): Prediction with detected bounding boxes and segmentation masks in form of sv.Detections(...) object.model_id(roboflow_model_id): Roboflow model id.
Example JSON definition of step Instance Segmentation Model in version v3
{
"name": "<your_step_name_here>",
"type": "roboflow_core/roboflow_instance_segmentation_model@v3",
"images": "$inputs.image",
"model_id": "my_project/3",
"confidence_mode": "<block_does_not_provide_example>",
"custom_confidence": 0.3,
"class_filter": [
"a",
"b",
"c"
],
"iou_threshold": 0.4,
"max_detections": 300,
"class_agnostic_nms": true,
"max_candidates": 3000,
"mask_decode_mode": "accurate",
"tradeoff_factor": 0.3,
"disable_active_learning": true,
"active_learning_target_dataset": "my_project",
"enforce_dense_masks_in_inference_models": true
}
v2¶
Class: RoboflowInstanceSegmentationModelBlockV2 (there are multiple versions of this block)
Warning: This block has multiple versions. Please refer to the specific version for details. You can learn more about how versions work here: Versioning
Run inference on an instance segmentation model hosted on or uploaded to Roboflow.
You can query any model that is private to your account, or any public model available on Roboflow Universe.
You will need to set your Roboflow API key in your Inference environment to use this block. To learn more about setting your Roboflow API key, refer to the Inference documentation.
Type identifier¶
Use the following identifier in step "type" field: roboflow_core/roboflow_instance_segmentation_model@v2to 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 |
Roboflow model identifier.. | ✅ |
confidence |
float |
Confidence threshold for predictions.. | ✅ |
class_filter |
List[str] |
List of accepted classes. Classes must exist in the model's training set.. | ✅ |
iou_threshold |
float |
Minimum overlap threshold between boxes to combine them into a single detection, used in NMS. Learn more.. | ✅ |
max_detections |
int |
Maximum number of detections to return.. | ✅ |
class_agnostic_nms |
bool |
Boolean flag to specify if NMS is to be used in class-agnostic mode.. | ✅ |
max_candidates |
int |
Maximum number of candidates as NMS input to be taken into account.. | ✅ |
mask_decode_mode |
str |
Parameter of mask decoding in prediction post-processing.. | ✅ |
tradeoff_factor |
float |
Post-processing parameter to dictate tradeoff between fast and accurate.. | ✅ |
disable_active_learning |
bool |
Boolean flag to disable project-level active learning for this block.. | ✅ |
active_learning_target_dataset |
str |
Target dataset for active learning, if enabled.. | ✅ |
enforce_dense_masks_in_inference_models |
bool |
Boolean flag to enforce dense masks when inference models backend is in use (irrelevant in other cases). Dense masks are faster to process, but require more memory. Users can't tweak this flag when running on Roboflow serverless platform.. | ✅ |
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 Instance Segmentation Model in version v2.
- inputs:
Template Matching,Morphological Transformation,Classification Label Visualization,Crop Visualization,Stability AI Outpainting,Blur Visualization,Reference Path Visualization,OpenAI,Anthropic Claude,Camera Focus,Instance Segmentation Model,Size Measurement,Model Comparison Visualization,Florence-2 Model,Trace Visualization,Background Color Visualization,JSON Parser,Label Visualization,Image Convert Grayscale,Florence-2 Model,Text Display,Qwen-VL,Llama 3.2 Vision,Image Blur,Keypoint Detection Model,Absolute Static Crop,CSV Formatter,Keypoint Detection Model,LMM,Qwen 3.5 API,Qwen 3.6 API,Camera Focus,Line Counter,VLM As Detector,Multi-Label Classification Model,Clip Comparison,Google Gemma API,Contrast Enhancement,Halo Visualization,Color Visualization,Morphological Transformation,MoonshotAI Kimi,Stitch OCR Detections,Event Writer,Buffer,Stability AI Inpainting,Roboflow Asset Library Attributes,Microsoft SQL Server Sink,OpenAI,Roboflow Vision Events,Identify Outliers,CogVLM,Detections Consensus,Object Detection Model,OPC UA Writer Sink,Semantic Segmentation Model,Dynamic Crop,Bounding Box Visualization,Qwen3.5-VL,Clip Comparison,OpenAI,SIFT Comparison,OCR Model,Single-Label Classification Model,Slack Notification,OpenRouter,Detection Event Log,SIFT Comparison,Pixelate Visualization,Google Vision OCR,Dynamic Zone,Google Gemma,Halo Visualization,Stitch OCR Detections,GLM-OCR,Image Threshold,Stitch Images,Twilio SMS/MMS Notification,Icon Visualization,VLM As Classifier,MoonshotAI Kimi,Single-Label Classification Model,Google Gemini,Single-Label Classification Model,Webhook Sink,Instance Segmentation Model,QR Code Generator,MQTT Writer,Ellipse Visualization,Object Detection Model,Keypoint Detection Model,Dot Visualization,Perspective Correction,Instance Segmentation Model,Roboflow Dataset Upload,PLC ModbusTCP,SIFT,Google Gemini,Dimension Collapse,EasyOCR,Local File Sink,Triangle Visualization,Contrast Equalization,Polygon Visualization,OpenAI,Heatmap Visualization,Detections List Roll-Up,Google Gemini,PLC EthernetIP,LMM For Classification,VLM As Detector,Identify Changes,Multi-Label Classification Model,Polygon Visualization,Image Stack,Llama 3.2 Vision,Mask Visualization,Anthropic Claude,Email Notification,Distance Measurement,PTZ Tracking (ONVIF),Keypoint Visualization,Background Subtraction,Multi-Label Classification Model,Twilio SMS Notification,Email Notification,Semantic Segmentation Model,Image Slicer,Image Contours,Line Counter Visualization,Image Preprocessing,VLM As Classifier,Depth Estimation,Pixel Color Count,Motion Detection,Current Time,Corner Visualization,Polygon Zone Visualization,Camera Calibration,Roboflow Dataset Upload,Grid Visualization,Stability AI Image Generation,S3 Sink,Circle Visualization,Image Slicer,Roboflow Custom Metadata,Relative Static Crop,Instance Segmentation Model,Model Monitoring Inference Aggregator,OpenAI-Compatible LLM,Object Detection Model,Anthropic Claude,Line Counter - outputs:
Halo Visualization,Overlap Analysis,GLM-OCR,SAM 3 Interactive,Crop Visualization,Icon Visualization,Detections Transformation,Blur Visualization,ByteTrack Tracker,Detections Classes Replacement,Single-Label Classification Model,Byte Tracker,Single-Label Classification Model,Webhook Sink,Instance Segmentation Model,Track Class Lock,Size Measurement,Mask Edge Snap,Instance Segmentation Model,Model Comparison Visualization,Path Deviation,Florence-2 Model,Trace Visualization,Ellipse Visualization,Object Detection Model,SmolVLM2,Keypoint Detection Model,BoT-SORT Tracker,Dot Visualization,Perspective Correction,Label Visualization,Instance Segmentation Model,Florence-2 Model,Per-Class Confidence Filter,Qwen-VL,Roboflow Dataset Upload,Detections Stabilizer,Keypoint Detection Model,Detections Merge,Velocity,Keypoint Detection Model,OC-SORT Tracker,Qwen2.5-VL,SAM 3,Triangle Visualization,Camera Focus,Time in Zone,Line Counter,SORT Tracker,SAM2 Video Tracker,Polygon Visualization,Qwen3-VL,Heatmap Visualization,Multi-Label Classification Model,Detections Stitch,Detections List Roll-Up,Halo Visualization,Color Visualization,Event Writer,Multi-Label Classification Model,Polygon Visualization,Mask Visualization,Detections Filter,Distance Measurement,Stability AI Inpainting,Bounding Rectangle,PTZ Tracking (ONVIF),Time in Zone,Overlap Filter,Roboflow Vision Events,Multi-Label Classification Model,Mask Area Measurement,Semantic Segmentation Model,Detection Offset,Detections Consensus,Object Detection Model,Byte Tracker,SAM 3,Semantic Segmentation Model,Dynamic Crop,Path Deviation,Byte Tracker,Bounding Box Visualization,Detections Combine,Qwen3.5-VL,Qwen3.5,Roboflow Dataset Upload,Corner Visualization,Moondream2,Segment Anything 2 Model,SAM 3,Circle Visualization,Time in Zone,Single-Label Classification Model,Roboflow Custom Metadata,Instance Segmentation Model,Model Monitoring Inference Aggregator,Object Detection Model,Detection Event Log,Pixelate Visualization,Background Color Visualization,Line Counter,SAM3 Video Tracker,Dynamic Zone
Input and Output Bindings¶
The available connections depend on its binding kinds. Check what binding kinds
Instance Segmentation Model in version v2 has.
Bindings
-
input
images(image): The image to infer on..model_id(roboflow_model_id): Roboflow model identifier..confidence(float_zero_to_one): Confidence threshold for predictions..class_filter(list_of_values): List of accepted classes. Classes must exist in the model's training set..iou_threshold(float_zero_to_one): Minimum overlap threshold between boxes to combine them into a single detection, used in NMS. Learn more..max_detections(integer): Maximum number of detections to return..class_agnostic_nms(boolean): Boolean flag to specify if NMS is to be used in class-agnostic mode..max_candidates(integer): Maximum number of candidates as NMS input to be taken into account..mask_decode_mode(string): Parameter of mask decoding in prediction post-processing..tradeoff_factor(float_zero_to_one): Post-processing parameter to dictate tradeoff between fast and accurate..disable_active_learning(boolean): Boolean flag to disable project-level active learning for this block..active_learning_target_dataset(roboflow_project): Target dataset for active learning, if enabled..enforce_dense_masks_in_inference_models(boolean): Boolean flag to enforce dense masks when inference models backend is in use (irrelevant in other cases). Dense masks are faster to process, but require more memory. Users can't tweak this flag when running on Roboflow serverless platform..
-
output
inference_id(inference_id): Inference identifier.predictions(instance_segmentation_prediction): Prediction with detected bounding boxes and segmentation masks in form of sv.Detections(...) object.model_id(roboflow_model_id): Roboflow model id.
Example JSON definition of step Instance Segmentation Model in version v2
{
"name": "<your_step_name_here>",
"type": "roboflow_core/roboflow_instance_segmentation_model@v2",
"images": "$inputs.image",
"model_id": "my_project/3",
"confidence": 0.3,
"class_filter": [
"a",
"b",
"c"
],
"iou_threshold": 0.4,
"max_detections": 300,
"class_agnostic_nms": true,
"max_candidates": 3000,
"mask_decode_mode": "accurate",
"tradeoff_factor": 0.3,
"disable_active_learning": true,
"active_learning_target_dataset": "my_project",
"enforce_dense_masks_in_inference_models": true
}
v1¶
Class: RoboflowInstanceSegmentationModelBlockV1 (there are multiple versions of this block)
Warning: This block has multiple versions. Please refer to the specific version for details. You can learn more about how versions work here: Versioning
Run inference on an instance segmentation model hosted on or uploaded to Roboflow.
You can query any model that is private to your account, or any public model available on Roboflow Universe.
You will need to set your Roboflow API key in your Inference environment to use this block. To learn more about setting your Roboflow API key, refer to the Inference documentation.
Type identifier¶
Use the following identifier in step "type" field: roboflow_core/roboflow_instance_segmentation_model@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 |
Roboflow model identifier.. | ✅ |
confidence |
float |
Confidence threshold for predictions.. | ✅ |
class_filter |
List[str] |
List of accepted classes. Classes must exist in the model's training set.. | ✅ |
iou_threshold |
float |
Minimum overlap threshold between boxes to combine them into a single detection, used in NMS. Learn more.. | ✅ |
max_detections |
int |
Maximum number of detections to return.. | ✅ |
class_agnostic_nms |
bool |
Boolean flag to specify if NMS is to be used in class-agnostic mode.. | ✅ |
max_candidates |
int |
Maximum number of candidates as NMS input to be taken into account.. | ✅ |
mask_decode_mode |
str |
Parameter of mask decoding in prediction post-processing.. | ✅ |
tradeoff_factor |
float |
Post-processing parameter to dictate tradeoff between fast and accurate.. | ✅ |
disable_active_learning |
bool |
Boolean flag to disable project-level active learning for this block.. | ✅ |
active_learning_target_dataset |
str |
Target dataset for active learning, if enabled.. | ✅ |
enforce_dense_masks_in_inference_models |
bool |
Boolean flag to enforce dense masks when inference models backend is in use (irrelevant in other cases). Dense masks are faster to process, but require more memory. Users can't tweak this flag when running on Roboflow serverless platform.. | ✅ |
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 Instance Segmentation Model in version v1.
- inputs:
Template Matching,Morphological Transformation,Classification Label Visualization,Crop Visualization,Stability AI Outpainting,Blur Visualization,Reference Path Visualization,OpenAI,Anthropic Claude,Camera Focus,Instance Segmentation Model,Size Measurement,Model Comparison Visualization,Florence-2 Model,Trace Visualization,Background Color Visualization,JSON Parser,Label Visualization,Image Convert Grayscale,Florence-2 Model,Text Display,Qwen-VL,Llama 3.2 Vision,Image Blur,Keypoint Detection Model,Absolute Static Crop,CSV Formatter,Keypoint Detection Model,LMM,Qwen 3.5 API,Qwen 3.6 API,Camera Focus,Line Counter,VLM As Detector,Multi-Label Classification Model,Clip Comparison,Google Gemma API,Contrast Enhancement,Halo Visualization,Color Visualization,Morphological Transformation,MoonshotAI Kimi,Stitch OCR Detections,Event Writer,Buffer,Stability AI Inpainting,Roboflow Asset Library Attributes,Microsoft SQL Server Sink,OpenAI,Roboflow Vision Events,Identify Outliers,CogVLM,Detections Consensus,Object Detection Model,OPC UA Writer Sink,Semantic Segmentation Model,Dynamic Crop,Bounding Box Visualization,Qwen3.5-VL,Clip Comparison,OpenAI,SIFT Comparison,OCR Model,Single-Label Classification Model,Slack Notification,OpenRouter,Detection Event Log,SIFT Comparison,Pixelate Visualization,Google Vision OCR,Dynamic Zone,Google Gemma,Halo Visualization,Stitch OCR Detections,GLM-OCR,Image Threshold,Stitch Images,Twilio SMS/MMS Notification,Icon Visualization,VLM As Classifier,MoonshotAI Kimi,Single-Label Classification Model,Google Gemini,Single-Label Classification Model,Webhook Sink,Instance Segmentation Model,QR Code Generator,MQTT Writer,Ellipse Visualization,Object Detection Model,Keypoint Detection Model,Dot Visualization,Perspective Correction,Instance Segmentation Model,Roboflow Dataset Upload,PLC ModbusTCP,SIFT,Google Gemini,Dimension Collapse,EasyOCR,Local File Sink,Triangle Visualization,Contrast Equalization,Polygon Visualization,OpenAI,Heatmap Visualization,Detections List Roll-Up,Google Gemini,PLC EthernetIP,LMM For Classification,VLM As Detector,Identify Changes,Multi-Label Classification Model,Polygon Visualization,Image Stack,Llama 3.2 Vision,Mask Visualization,Anthropic Claude,Email Notification,Distance Measurement,PTZ Tracking (ONVIF),Keypoint Visualization,Background Subtraction,Multi-Label Classification Model,Twilio SMS Notification,Email Notification,Semantic Segmentation Model,Image Slicer,Image Contours,Line Counter Visualization,Image Preprocessing,VLM As Classifier,Depth Estimation,Pixel Color Count,Motion Detection,Current Time,Corner Visualization,Polygon Zone Visualization,Camera Calibration,Roboflow Dataset Upload,Grid Visualization,Stability AI Image Generation,S3 Sink,Circle Visualization,Image Slicer,Roboflow Custom Metadata,Relative Static Crop,Instance Segmentation Model,Model Monitoring Inference Aggregator,OpenAI-Compatible LLM,Object Detection Model,Anthropic Claude,Line Counter - outputs:
Overlap Analysis,Morphological Transformation,Classification Label Visualization,Crop Visualization,Stability AI Outpainting,Detections Transformation,Blur Visualization,Reference Path Visualization,OpenAI,YOLO-World Model,Detections Classes Replacement,Anthropic Claude,Track Class Lock,Instance Segmentation Model,Size Measurement,Mask Edge Snap,Model Comparison Visualization,Florence-2 Model,Trace Visualization,Label Visualization,Florence-2 Model,Qwen-VL,Text Display,Llama 3.2 Vision,Image Blur,Velocity,Keypoint Detection Model,LMM,OC-SORT Tracker,Qwen 3.5 API,Qwen 3.6 API,Camera Focus,Line Counter,SORT Tracker,Detections Stitch,Clip Comparison,Google Gemma API,Halo Visualization,Stitch OCR Detections,MoonshotAI Kimi,Color Visualization,Morphological Transformation,Event Writer,Stability AI Inpainting,Cache Set,Bounding Rectangle,Microsoft SQL Server Sink,Time in Zone,Roboflow Asset Library Attributes,OpenAI,Roboflow Vision Events,Mask Area Measurement,Detection Offset,CogVLM,Detections Consensus,OPC UA Writer Sink,Semantic Segmentation Model,Dynamic Crop,Path Deviation,Byte Tracker,Bounding Box Visualization,Detections Combine,Qwen3.5-VL,SAM 3,Cache Get,OpenAI,Time in Zone,Slack Notification,OpenRouter,Detection Event Log,Google Vision OCR,SIFT Comparison,Pixelate Visualization,SAM3 Video Tracker,Google Gemma,Dynamic Zone,Halo Visualization,CLIP Embedding Model,Stitch OCR Detections,GLM-OCR,Image Threshold,SAM 3 Interactive,Twilio SMS/MMS Notification,Icon Visualization,MoonshotAI Kimi,ByteTrack Tracker,Single-Label Classification Model,Google Gemini,Byte Tracker,Webhook Sink,Instance Segmentation Model,QR Code Generator,Path Deviation,MQTT Writer,Ellipse Visualization,Anthropic Claude,BoT-SORT Tracker,Dot Visualization,Perspective Correction,Instance Segmentation Model,Seg Preview,Per-Class Confidence Filter,Roboflow Dataset Upload,Detections Stabilizer,Detections Merge,Google Gemini,Local File Sink,SAM 3,Triangle Visualization,Contrast Equalization,Time in Zone,Polygon Visualization,OpenAI,SAM2 Video Tracker,Heatmap Visualization,Perception Encoder Embedding Model,Detections List Roll-Up,Google Gemini,LMM For Classification,Llama 3.2 Vision,Multi-Label Classification Model,Polygon Visualization,Email Notification,Mask Visualization,Anthropic Claude,Detections Filter,Distance Measurement,PTZ Tracking (ONVIF),Keypoint Visualization,Overlap Filter,Twilio SMS Notification,Email Notification,Line Counter Visualization,Image Preprocessing,SAM 3,Byte Tracker,Depth Estimation,Pixel Color Count,Current Time,Roboflow Dataset Upload,Polygon Zone Visualization,Moondream2,Segment Anything 2 Model,Corner Visualization,Stability AI Image Generation,S3 Sink,Circle Visualization,Roboflow Custom Metadata,Instance Segmentation Model,Model Monitoring Inference Aggregator,OpenAI-Compatible LLM,Object Detection Model,Background Color Visualization,Line Counter
Input and Output Bindings¶
The available connections depend on its binding kinds. Check what binding kinds
Instance Segmentation Model in version v1 has.
Bindings
-
input
images(image): The image to infer on..model_id(roboflow_model_id): Roboflow model identifier..confidence(float_zero_to_one): Confidence threshold for predictions..class_filter(list_of_values): List of accepted classes. Classes must exist in the model's training set..iou_threshold(float_zero_to_one): Minimum overlap threshold between boxes to combine them into a single detection, used in NMS. Learn more..max_detections(integer): Maximum number of detections to return..class_agnostic_nms(boolean): Boolean flag to specify if NMS is to be used in class-agnostic mode..max_candidates(integer): Maximum number of candidates as NMS input to be taken into account..mask_decode_mode(string): Parameter of mask decoding in prediction post-processing..tradeoff_factor(float_zero_to_one): Post-processing parameter to dictate tradeoff between fast and accurate..disable_active_learning(boolean): Boolean flag to disable project-level active learning for this block..active_learning_target_dataset(roboflow_project): Target dataset for active learning, if enabled..enforce_dense_masks_in_inference_models(boolean): Boolean flag to enforce dense masks when inference models backend is in use (irrelevant in other cases). Dense masks are faster to process, but require more memory. Users can't tweak this flag when running on Roboflow serverless platform..
-
output
inference_id(string): String value.predictions(instance_segmentation_prediction): Prediction with detected bounding boxes and segmentation masks in form of sv.Detections(...) object.
Example JSON definition of step Instance Segmentation Model in version v1
{
"name": "<your_step_name_here>",
"type": "roboflow_core/roboflow_instance_segmentation_model@v1",
"images": "$inputs.image",
"model_id": "my_project/3",
"confidence": 0.3,
"class_filter": [
"a",
"b",
"c"
],
"iou_threshold": 0.4,
"max_detections": 300,
"class_agnostic_nms": true,
"max_candidates": 3000,
"mask_decode_mode": "accurate",
"tradeoff_factor": 0.3,
"disable_active_learning": true,
"active_learning_target_dataset": "my_project",
"enforce_dense_masks_in_inference_models": true
}