canny-chenn 1.0.5

A simple Canny edge detector for pytorch

Canny Edges

 

A simple Torch.nn.Module to return an image mask representing edges found by the Canny Edge-Finding algorithm.

Supports:

• Linux
• PyTorch Tensor images formatted in image batches [B, C, h, w], with float32 data, and values spanning [0,1]
• Any number of image channels C
• All PyTorch backends
• TorchScript jit script, as well as the newer, traced torch-export, on the Canny nn.Module class

pip install

pip install canny_chenn
# or,
pixi add --pypi canny_chenn
# or,
uv add canny_chenn

then

from torch import Tensor
from canny_chenn import Canny
canny = Canny()
img_batch = Tensor([[[[0, 1, 2, 3], [0, 1, 2, 3], [0, 1, 2, 3]],
            [[0, 1, 2, 3], [0, 1, 2, 3], [0, 1, 2, 3]],
            [[0, 1, 2, 3], [0, 1, 2, 3], [0, 1, 2, 3]],
            [[0, 1, 2, 3], [0, 1, 2, 3], [0, 1, 2, 3]]]])
edge_mask = canny(img_batch)
# tensor([[[[0.0000, 0.6333, 0.0000, 0.0000],
#           [0.0000, 0.6333, 0.0000, 0.0000],
#           [0.0000, 0.6333, 0.0000, 0.0000]]]])

---

Dev

I use mise to handle project tools.

# clone, install tools
git clone https://github.com/chrishenn/canny.git
cd canny
mise i

# list available just recipes
just --list

Available recipes:
    demo   # Show canny detections on a set of test images. Kill with ctrl+c
    lint   # run project formatters/linters [alias: l]
    test   # run basic tests
    unsafe # run ruff with unsafe-fixes
    
# build wheel, sdist pkgs locally
uv build --package canny_chenn

# push tag to trigger github action to build pkgs and publish to pypi
git tag -a v1.0.4 -m v1.0.4 -f && git push --tags -f