Controlnet cli tools

Simple tools to generate controlnet mask images using OpenCV:

• canny
• depth (fake depth)
• eye_mask
• face_mask
• laplace
• sobel
• soft edge

Deps: opencv-dev

BUILD

Build with cmake in build/ directory, then run binaries from build/.

cmake . -B build
cmake --build build

CLI Tools

The project provides several CLI tools for image processing (edge detection, masks). All tools use unified arguments: -i/--input and -o/--output are required. Run with --help for full usage.

canny_cli

Canny edge detection.

Flags:

• -i, --input: String, required, input image.
• -o, --output: String, required, output edge map.
• -l, --low-threshold: int, optional, default 100, Canny low threshold.
• -h, --high-threshold: int, optional, default 200, Canny high threshold.
• -b, --blur: int, optional, default 0, blur kernel size.
• --help: Print usage.

Example:

./build/canny_cli -i input.jpg -o edges.png -b 3 -l 50 --high-threshold 150

depth_cli

Fake depth map generation.

Flags:

• -i, --input: String, required.
• -o, --output: String, required.
• -b, --blur: int, optional, default 0.
• --help: Print usage.

Example:

./build/depth_cli --input input.jpg -o depth.png -b 1

eye_mask_cli

Eye detection and masking.

Flags:

• -i, --input: String, required.
• -o, --output: String, required.
• -s, --scale: float, optional, default 1.1, detection scale.
• -n, --min-neighbors: int, optional, default 3.
• -z, --min-size: int, optional, default 30.
• --help: Print usage. (Also supports -h for help.)

Example:

./build/eye_mask_cli -i face.jpg --output eyes_mask.png -s 1.2 -n 5 -z 40

face_mask_cli

Face detection and masking.

Flags:

• -i, --input: String, required.
• -o, --output: String, required.
• -s, --scale: float, optional, default 1.1.
• -n, --min-neighbors: int, optional, default 3.
• -z, --min-size: int, optional, default 80.
• -c, --cascade: string, optional, default built-in, cascade file.
• -p, --shape: string, optional, default "oval", shape (oval/rectangle).
• --help: Print usage.

Example:

./build/face_mask_cli -i group.jpg -o faces_mask.png --shape rectangle -c haarcascade_frontalface_alt.xml

laplace_cli

Laplace edge detection.

Flags:

• -i, --input: String, required.
• -o, --output: String, required.
• -k, --ksize: int, optional, default 3, kernel size.
• -b, --blur: int, optional, default 0.
• --help: Print usage.

Example:

./build/laplace_cli --input img.jpg -o laplace_edges.png -k 5 -b 0

sobel_cli

Sobel edge detection.

Flags:

• -i, --input: String, required.
• -o, --output: String, required.
• -x, --dx: int, optional, default 1, dx order.
• -y, --dy: int, optional, default 1, dy order.
• -k, --ksize: int, optional, default 3, kernel size.
• -b, --blur: int, optional, default 0.
• --scale: float, optional, default 1.0, scale factor.
• --delta: float, optional, default 0.0, delta value.
• --help: Print usage.

Example:

./build/sobel_cli -i img.jpg --output sobel.png --dx 1 --dy 1 -k 3

soft_edge_cli

Soft edge detection using Difference of Gaussians.

Flags:

• -i, --input: String, required.
• -o, --output: String, required.
• -b, --blur: int, optional, default 0.
• --sigma: float, optional, default 1.0, sigma for DoG.
• --help: Print usage.

Example:

./build/soft_edge_cli -i soft.jpg -o soft_edges.png -b 7 --sigma 2.0

For controlnet models see: https://huggingface.co/comfyanonymous/ControlNet-v1-1_fp16_safetensors/tree/main stable diffusion cpp compatible models: https://huggingface.co/lllyasviel/ControlNet-v1-1/tree/459bf90295ac305bc3ae8266e39a089f433eab4f