Crop Visualization

Class: CropVisualizationBlockV1

Source: inference.core.workflows.core_steps.visualizations.crop.v1.CropVisualizationBlockV1

Display scaled-up, zoomed-in views of detected objects overlaid on the original image, allowing detailed inspection of small or distant objects while maintaining context with the full scene.

How This Block Works

This block takes an image and detection predictions and creates scaled-up, zoomed-in crops of each detected object, then displays these enlarged crops on the original image. The block:

1. Takes an image and predictions as input
2. Identifies detected regions from bounding boxes or segmentation masks
3. Extracts the image region for each detected object (crops the object from the original image)
4. Scales up each crop by the specified scale factor (e.g., 2x makes objects twice as large)
5. Applies color styling to the crop border based on the selected color palette, with colors assigned by class, index, or track ID
6. Positions the scaled crop on the image at the specified anchor point relative to the original detection location using Supervision's CropAnnotator
7. Draws a colored border around the scaled crop with the specified thickness
8. Returns an annotated image with scaled-up object crops overlaid on the original image

The block works with both object detection predictions (using bounding boxes) and instance segmentation predictions (using masks). When masks are available, it crops the exact shape of detected objects; otherwise, it crops rectangular bounding box regions. The scale factor allows you to zoom in on objects, making small or distant objects more visible and easier to inspect. The scaled crops are positioned relative to their original detection locations, allowing you to see both the zoomed-in detail and the object's position in the full scene context.

Common Use Cases

• Small Object Inspection: Zoom in on small detected objects (e.g., defects, small products, distant objects) to make them more visible and easier to inspect while maintaining scene context
• Detail Visualization: Display enlarged views of detected objects for detailed analysis, quality control, or inspection workflows where fine details need to be visible
• Multi-Scale Object Display: Show both the full scene and zoomed-in object details simultaneously, useful for applications where context and detail are both important
• Quality Control and Inspection: Inspect detected defects, products, or components at higher magnification while keeping the original detection location visible for reference
• Presentation and Reporting: Create visualizations that highlight detected objects with zoomed-in views for reports, documentation, or presentations where both overview and detail are needed
• User Interface Enhancement: Provide zoomed-in object views in user interfaces, dashboards, or interactive applications where users need to see object details without losing scene context

Connecting to Other Blocks

The annotated image from this block can be connected to:

• Other visualization blocks (e.g., Label Visualization, Bounding Box Visualization, Polygon Visualization) to combine scaled crops with additional annotations for comprehensive visualization
• Data storage blocks (e.g., Local File Sink, CSV Formatter, Roboflow Dataset Upload) to save images with scaled crops for documentation, reporting, or analysis
• Webhook blocks to send visualized results with scaled crops to external systems, APIs, or web applications for display in dashboards or monitoring tools
• Notification blocks (e.g., Email Notification, Slack Notification) to send annotated images with scaled crops as visual evidence in alerts or reports
• Video output blocks to create annotated video streams or recordings with scaled crops for live monitoring, detailed inspection, or post-processing analysis

Type identifier

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

Properties

Name	Type	Description	Refs
name	str	Enter a unique identifier for this step..	❌
copy_image	bool	Enable this option to create a copy of the input image for visualization, preserving the original. Use this when stacking multiple visualizations..	✅
color_palette	str	Select a color palette for the visualised elements..	✅
palette_size	int	Specify the number of colors in the palette. This applies when using custom or Matplotlib palettes..	✅
custom_colors	List[str]	Define a list of custom colors for bounding boxes in HEX format..	✅
color_axis	str	Choose how bounding box colors are assigned..	✅
position	str	Anchor position for placing the scaled crop relative to the original detection's bounding box. Options include: CENTER (center of box), corners (TOP_LEFT, TOP_RIGHT, BOTTOM_LEFT, BOTTOM_RIGHT), edge midpoints (TOP_CENTER, CENTER_LEFT, CENTER_RIGHT, BOTTOM_CENTER), or CENTER_OF_MASS (center of mass of the object). The scaled crop will be positioned at this anchor point relative to the original detection location..	✅
scale_factor	float	Factor by which to scale (zoom) the cropped object region. A factor of 2.0 doubles the size of the crop, making objects twice as large. A factor of 1.0 shows the crop at original size. Higher values (e.g., 3.0, 4.0) create more zoomed-in views, useful for inspecting small or distant objects. Lower values (e.g., 1.5) provide subtle magnification..	✅
border_thickness	int	Thickness of the border outline around the scaled crop in pixels. Higher values create thicker, more visible borders that help distinguish the scaled crop from the background..	✅

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 Crop Visualization in version v1.

• inputs: OpenAI, Stitch OCR Detections, VLM As Detector, Distance Measurement, VLM As Classifier, Depth Estimation, Time in Zone, Camera Focus, Email Notification, Motion Detection, Classification Label Visualization, Velocity, Polygon Visualization, Size Measurement, Cosine Similarity, Template Matching, Roboflow Dataset Upload, Corner Visualization, Time in Zone, JSON Parser, Object Detection Model, Byte Tracker, Detection Offset, Reference Path Visualization, LMM For Classification, Local File Sink, EasyOCR, Stability AI Outpainting, Halo Visualization, SAM 3, Grid Visualization, Polygon Visualization, Florence-2 Model, VLM As Detector, CogVLM, Gaze Detection, Anthropic Claude, Heatmap Visualization, Background Subtraction, SIFT Comparison, SIFT Comparison, Image Contours, Roboflow Custom Metadata, Label Visualization, Stitch OCR Detections, OpenAI, Stability AI Image Generation, Dynamic Zone, Google Gemini, VLM As Classifier, Detections Transformation, Bounding Box Visualization, Blur Visualization, Image Preprocessing, Llama 3.2 Vision, Relative Static Crop, Icon Visualization, Clip Comparison, Keypoint Detection Model, Byte Tracker, Pixelate Visualization, Halo Visualization, Image Blur, Webhook Sink, Detections List Roll-Up, Byte Tracker, Line Counter, Stitch Images, Time in Zone, Trace Visualization, Crop Visualization, Object Detection Model, Google Gemini, Detections Stabilizer, SAM 3, Instance Segmentation Model, Multi-Label Classification Model, Ellipse Visualization, Buffer, Twilio SMS Notification, Bounding Rectangle, SAM 3, PTZ Tracking (ONVIF).md), Detections Classes Replacement, Morphological Transformation, Moondream2, SIFT, Triangle Visualization, Detection Event Log, Dot Visualization, Model Comparison Visualization, Slack Notification, Clip Comparison, Contrast Equalization, Mask Visualization, Google Vision OCR, Circle Visualization, Dynamic Crop, Email Notification, Perspective Correction, Line Counter, YOLO-World Model, OpenAI, Roboflow Dataset Upload, Text Display, Identify Outliers, OpenAI, Line Counter Visualization, Polygon Zone Visualization, Overlap Filter, Segment Anything 2 Model, QR Code Generator, Detections Combine, CSV Formatter, Absolute Static Crop, Detections Stitch, Google Gemini, Detections Filter, Twilio SMS/MMS Notification, Image Threshold, Pixel Color Count, Anthropic Claude, Identify Changes, Camera Calibration, Model Monitoring Inference Aggregator, Detections Merge, Anthropic Claude, Image Slicer, LMM, Instance Segmentation Model, Single-Label Classification Model, Background Color Visualization, Florence-2 Model, Path Deviation, Keypoint Visualization, Keypoint Detection Model, Image Convert Grayscale, Seg Preview, Camera Focus, Mask Area Measurement, Stability AI Inpainting, Dimension Collapse, Color Visualization, Path Deviation, Detections Consensus, Image Slicer, OCR Model
• outputs: Object Detection Model, OpenAI, QR Code Detection, Google Gemini, SAM 3, Detections Stabilizer, Instance Segmentation Model, VLM As Detector, Multi-Label Classification Model, Ellipse Visualization, Buffer, VLM As Classifier, Depth Estimation, Camera Focus, Email Notification, SAM 3, Motion Detection, Morphological Transformation, Classification Label Visualization, Moondream2, Polygon Visualization, Template Matching, SIFT, Triangle Visualization, Roboflow Dataset Upload, Corner Visualization, Dot Visualization, Model Comparison Visualization, Object Detection Model, Clip Comparison, Contrast Equalization, LMM For Classification, Reference Path Visualization, Qwen2.5-VL, EasyOCR, Mask Visualization, Google Vision OCR, Stability AI Outpainting, SAM 3, Single-Label Classification Model, Halo Visualization, Circle Visualization, Dynamic Crop, Stability AI Inpainting, Perspective Correction, Polygon Visualization, OpenAI, YOLO-World Model, Florence-2 Model, Roboflow Dataset Upload, Text Display, VLM As Detector, CogVLM, Gaze Detection, Anthropic Claude, Trace Visualization, Background Subtraction, Heatmap Visualization, OpenAI, Line Counter Visualization, Polygon Zone Visualization, SIFT Comparison, Dominant Color, Segment Anything 2 Model, SmolVLM2, Perception Encoder Embedding Model, Image Contours, Absolute Static Crop, Google Gemini, Label Visualization, Detections Stitch, OpenAI, Stability AI Image Generation, Qwen3-VL, Twilio SMS/MMS Notification, Multi-Label Classification Model, Image Threshold, Google Gemini, Pixel Color Count, VLM As Classifier, Anthropic Claude, Bounding Box Visualization, Blur Visualization, Image Preprocessing, Llama 3.2 Vision, Camera Calibration, Anthropic Claude, Relative Static Crop, Image Slicer, Icon Visualization, Clip Comparison, LMM, Keypoint Detection Model, Byte Tracker, Instance Segmentation Model, CLIP Embedding Model, Single-Label Classification Model, Pixelate Visualization, Halo Visualization, Background Color Visualization, Image Blur, Florence-2 Model, Keypoint Visualization, Keypoint Detection Model, Image Convert Grayscale, Stitch Images, Seg Preview, Camera Focus, Time in Zone, Color Visualization, Barcode Detection, Crop Visualization, Image Slicer, OCR Model

Input and Output Bindings

The available connections depend on its binding kinds. Check what binding kinds Crop Visualization in version v1 has.

Bindings

• input

  • image (image): The image to visualize on..
  • copy_image (boolean): Enable this option to create a copy of the input image for visualization, preserving the original. Use this when stacking multiple visualizations..
  • predictions (Union[object_detection_prediction, rle_instance_segmentation_prediction, instance_segmentation_prediction, keypoint_detection_prediction]): Model predictions to visualize..
  • color_palette (string): Select a color palette for the visualised elements..
  • palette_size (integer): Specify the number of colors in the palette. This applies when using custom or Matplotlib palettes..
  • custom_colors (list_of_values): Define a list of custom colors for bounding boxes in HEX format..
  • color_axis (string): Choose how bounding box colors are assigned..
  • position (string): Anchor position for placing the scaled crop relative to the original detection's bounding box. Options include: CENTER (center of box), corners (TOP_LEFT, TOP_RIGHT, BOTTOM_LEFT, BOTTOM_RIGHT), edge midpoints (TOP_CENTER, CENTER_LEFT, CENTER_RIGHT, BOTTOM_CENTER), or CENTER_OF_MASS (center of mass of the object). The scaled crop will be positioned at this anchor point relative to the original detection location..
  • scale_factor (float): Factor by which to scale (zoom) the cropped object region. A factor of 2.0 doubles the size of the crop, making objects twice as large. A factor of 1.0 shows the crop at original size. Higher values (e.g., 3.0, 4.0) create more zoomed-in views, useful for inspecting small or distant objects. Lower values (e.g., 1.5) provide subtle magnification..
  • border_thickness (integer): Thickness of the border outline around the scaled crop in pixels. Higher values create thicker, more visible borders that help distinguish the scaled crop from the background..

• output

  • image (image): Image in workflows.

Example JSON definition of step Crop Visualization in version v1

{
    "name": "<your_step_name_here>",
    "type": "roboflow_core/crop_visualization@v1",
    "image": "$inputs.image",
    "copy_image": true,
    "predictions": "$steps.object_detection_model.predictions",
    "color_palette": "DEFAULT",
    "palette_size": 10,
    "custom_colors": [
        "#FF0000",
        "#00FF00",
        "#0000FF"
    ],
    "color_axis": "CLASS",
    "position": "CENTER",
    "scale_factor": 2.0,
    "border_thickness": 2
}