API reference

One page per kornia module, grouped by what you are trying to do. Image operators take batched (B, C, H, W) float tensors in [0, 1] and follow the conventions (points, boxes, local features and camera matrices have their own layouts, documented on their pages); most functions also have an nn.Module counterpart.

Image processing

kornia.color	Color space conversions, color maps and Bayer RAW processing.
kornia.filters	Blurring, edge detection, thresholding and custom kernels.
kornia.enhance	Intensity adjustments, histogram equalization, normalization and a differentiable JPEG codec.
kornia.morphology	Dilation, erosion, opening, closing, gradient, top hat and bottom hat.

Geometry & 3D

kornia.geometry	Warps, camera models, conversions, epipolar geometry, Lie groups, RANSAC and more.
kornia.sensors	Experimental camera model API.

Training

kornia.augmentation	Random and deterministic augmentations for images, masks, boxes, keypoints and video, with transform tracking and inversion.
kornia.losses	Reconstruction, segmentation, distribution and morphology losses.
kornia.metrics	Classification, segmentation, detection, image quality, flow, stereo and pose metrics.

Features & matching

kornia.feature

Local feature detectors, descriptors and matchers, classical and learned.

Models

kornia.models	Builders for the pretrained detection, edge detection, segmentation and tracking models.
kornia.contrib	Experimental operators and model wrappers: face detection, object detection, visual prompting, image stitching, KMeans.
kornia.tracking	Homography tracking.

Data & deployment

kornia.io	Load and save images as tensors.
kornia.image	Image container, drawing, tensor/NumPy conversion.
kornia.onnx	Run and chain ONNX models with ONNX Runtime.
kornia.core	Tensor wrapper and shared utilities.