Annotator-CannyEdge — Annotation AI Model
Canny edge detection, classical OpenCV — white outlines on black, with low and high thresholds tuning what survives. No neural net, so it runs on CPU in under a second, and feeds ControlNet Canny directly.
Details
- CategoryAnnotation
- Year2022
- LicenseApache-2.0
Inputs & Outputs
- ImageInputimage
Source image to extract edges from. Works best on images with clear structural boundaries (architecture, objects, illustrations). RGB or RGBA accepted.
- Edge MapOutputimage
Canny edge map: white edges on black background, same resolution as input. Feed directly into ControlNet-XL-Canny or ControlNet-XL-Union (canny mode).
Parameters
- low_thresholdintegerdefault 100range 0–255
Low threshold (0-255)
Minimum gradient magnitude to consider as an edge candidate (hysteresis lower bound). Lower values detect more edges including weak ones. Default 100. Raise for cleaner, fewer edges on noisy inputs.
- high_thresholdintegerdefault 200range 0–255
High threshold (0-255)
Gradient magnitude above which a pixel is definitely an edge (hysteresis upper bound). Default 200. Edges between low and high thresholds are kept only if connected to a strong edge. Lower for more aggressive edge inclusion.
Tags
Alternatives in Annotation
- Sticky Note
Canvas text annotation (no I/O).