Try OpenCV filters, then take the code with you.

Build an edge-detection recipe

1. Open an image

   Choose Open image, or drop a JPG, PNG, WebP or BMP into the preview. A clear photograph of a building, desk or printed page is a useful starting point.

2. Add Grayscale

   Choose Grayscale from the tools on the left. It appears in the Processing recipe on the right, and the preview updates.

3. Add Gaussian blur

   Choose Gaussian blur. Start with Kernel size 5 and Sigma 0. This smooths small details before edge detection. Click a recipe step to return to its settings.

4. Add Canny edges

   Switch to Advanced and choose Canny edges. Start with a low threshold of 50 and a high threshold of 150. Keep the low value at or below the high value.

5. Compare and adjust

   Use Show original to compare. Try increasing the thresholds if the result has too many edges. Turn a recipe step off with its checkbox to see its effect. The arrow buttons change the order of steps.

6. Save the image and recipe

   Choose Export image for the result. Choose Save recipe to download your settings as a JSON file; Load recipe brings them back later. Save before closing or reloading the tab.

7. Export the Python code

   Click Export Python code. The downloaded maple_image_recipe.py contains the enabled operations and their settings. It processes an input image and writes a new output file.

Run the exported code

If Python is installed, open a terminal in the folder containing the script and your input image. Install its dependencies, then run it with input and output filenames:

python -m pip install opencv-python numpy
python maple_image_recipe.py input.png edges.png

Use your own filename in place of input.png. The output file is edges.png. Open the script in a text editor to inspect the recipe and the OpenCV calls.

Reuse the recipe

Batch export (PNG ZIP) applies the same recipe to several images and downloads the results together. Check that the settings suit each image, especially when using crop or perspective coordinates.

What this edition covers

The web app has 51 operations, including adjustments, filters, thresholds, morphology and feature detection. It does not include every OpenCV function, trained AI models, RAW editing or painting layers. Images use 8-bit colour and are limited to 16 megapixels each. Python and browser results can differ slightly between OpenCV versions.