tinytex

Python texture sampling, processing and synthesis library for PyTorch-involved projects.

This library is a hodgepodge of tangentially-related procedures useful for sampling, creating and modifying various kinds of textures. This is primarily intended for batched or unbatched PyTorch image tensors. This library provides:

image resampling/rescaling, cropping and padding
tiling
- split images into tiles
- merge tiles back into images
- seamlessly stitch textures with color or vector data for mutual tiling or self-tiling
texture atlases
- pack images into texture atlases
- sample images from texture atlases
- generate tiling masks from texture atlases
computing and rendering 2D signed distance fields
computing and approximating surface geometry
- normals to height
- height to normals
- height/normals to curvature
approximating ambient occlusion and bent normals
blending multiple normal maps
pseudo-random number generation
generating tiling spatial-domain noise
generating spectral-domain noise
warping image coordinates
transforming 1D and 2D images to and from Haar wavelet coefficients
(experimental) backend-agnostic 1D/2D/3D textures for Taichi (if installed with Taichi optional dependency)
- load from and save to the filesystem
- convert textures to and from PyTorch tensors
- sample textures with lower or higher-order interpolation/approximation