Kornia Computer Vision Tutorials

GPU-accelerated computer vision tutorials built with Kornia and PyTorch.

Table of Contents

• Installation
• Quick Start
• Modules
• Running a Tutorial
• Pipeline CLI
• Testing
• Project Structure
• License

Installation

# Editable install (recommended for development)
pip install -e .

# Or with dev dependencies
pip install -e ".[dev]"

# Or minimal via requirements
pip install -r requirements.txt

Requires Python 3.10+, PyTorch 1.13+, and Kornia 0.6+.

Quick Start

import kornia
import torch

# Load and convert an image
img = kornia.image.image_to_tensor("tutorials/data/drslump.jpg")
tensor = img.float()[None, ...] / 255.0

# Apply operations on GPU if available
tensor = tensor.to("cuda" if torch.cuda.is_available() else "cpu")
blurred = kornia.filters.gaussian_blur2d(tensor, (11, 11), (2.0, 2.0))
edges = kornia.filters.sobel(kornia.color.rgb_to_grayscale(blurred))
print(f"Edges shape: {edges.shape}")

Modules

Module	Description
basic_operations/	Image I/O, flipping, resizing
filtering/	Box, Gaussian, median blur; unsharp masking
edge_detection/	Sobel, Laplacian, Canny, CNN-based, line segments, LSD
color_adjustments/	Brightness, contrast, saturation
feature_detection/	Gradients, Hough lines, face detection
benchmark/	Performance benchmarking (PSNR, timing comparison with OpenCV)
pipeline.py	CLI pipeline that chains multiple operations
custom_ops.py	Custom 2D convolution implementation
_utils.py	Shared utilities (loading, tensor conversion, CLI parsing)

Running a Tutorial

python tutorials/basic_operations/readimg.py <image_path>
python tutorials/filtering/blur.py <image_path> --kernel 11
python tutorials/edge_detection/edgedetection.py <image_path> --algo canny

Pipeline CLI

The pipeline chains blur, edge detection, and color adjustments in a single command:

python tutorials/pipeline.py input.jpg \
  --blur gaussian \
  --edges canny \
  --brightness 0.8 \
  --contrast 1.2 \
  --saturation 0.9 \
  --output result.jpg

Testing

# Run all tests
pytest tests/ -v

# Run with coverage
pip install pytest-cov
pytest tests/ --cov=tutorials --cov-report=html

The test suite covers:

• tests/test_imports.py — All modules import without errors
• tests/test_operations.py — Correctness of each operation (flip, resize, blur, edge detection, color adjustments)
• tests/test_all.py — Edge cases, property-based constraints, custom convolution, pipeline integration, utilities, and benchmarking

Project Structure

├── tutorials/
│   ├── __init__.py
│   ├── _utils.py              Shared image loading & tensor conversion
│   ├── custom_ops.py          Custom 2D convolution
│   ├── pipeline.py            CLI pipeline (blur → edges → color adj)
│   ├── basic_operations/      Image I/O, flipping, resizing
│   ├── filtering/             Blur techniques, unsharp mask
│   ├── edge_detection/        Edge & line detection
│   ├── color_adjustments/     Brightness, contrast, saturation
│   ├── feature_detection/     Gradients, Hough, face detection
│   ├── benchmark/             Performance benchmarking
│   └── data/                  Sample test images
├── tests/
│   ├── test_imports.py
│   ├── test_operations.py
│   └── test_all.py
├── docs/                      Sphinx documentation source
├── pyproject.toml
└── README.md

License

MIT