Build a Canny edge-detection workflow

Starter or above

Find object boundaries in a still image with Gaussian smoothing followed by Canny edge detection. This four-node workflow fits within Starter’s 15-node limit.

Workflow

Image Input → GaussianBlur → Canny → Image Output

Build it in Visionodo

  1. Create a free account, open Workspaces, create a workspace and wait for OpenCV to finish loading in the editor.
  2. Add Image Input, GaussianBlur, Canny and Image Output. Connect each output to the next input in the order shown above. Select a local image using the Image Input file control.
  3. Start with a small odd blur kernel, such as 5. Blur suppresses small variations that would otherwise appear as extra edges. Large kernels can erase fine details.
  4. Start with Canny’s low threshold at 50 and high threshold at 150. Run the workflow and inspect the edge image. Increase the thresholds if too much texture is detected; decrease them if useful boundaries disappear.
  5. Compare the same image with and without smoothing. Save the workflow to revisit the settings; use Save Image to download a result. The Image Output node is a preview and does not automatically download files.

What to expect

The output shows bright detected edges on a dark background. It identifies intensity boundaries, not object identities or semantic categories. Lighting, image size and noise affect the thresholds that work well.

Limits and practical checks

Media computation runs in your browser. Saving a graph can also save image data embedded in it; avoid embedding private images in cloud-saved workflows.

Start building free