fbcotter/pytorch_wavelets

Health v2 · maintenance only

Flags: no_license

Adoption not part of the score

1175 stars · 160 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-30, confidence not recorded

A PyTorch library implementing 2D discrete wavelet transforms (DWT), dual-tree complex wavelet transforms (DTCWT), their inverses, and a DTCWT-based ScatterNet, all with gradient support. It is designed for batches of multichannel images in NCHW format and also supports 1D DWT.

Use cases

• compute differentiable 2D wavelet transforms in pytorch
• run dual-tree complex wavelet transform on image batches on GPU
• build a fast DTCWT scatternet feature extractor for deep learning
• do 1D discrete wavelet decomposition of signals in pytorch
• invert wavelet transforms with gradient flow for training

When to choose

• you need wavelet transforms as differentiable layers inside a PyTorch model
• you want a faster alternative to KymatIO's Morlet scatternet
• you work with batches of multichannel images in NCHW format

When to avoid

• you need wavelets outside PyTorch (e.g. NumPy-only workflows)
• you need 3D or higher-dimensional wavelet transforms
• you need actively developed features beyond the 2023 release

Facets

library · maturity maintenance

image-processing machine-learning deep-learning deep-learning computer-vision image-processing machine-learning python cross-platform pytorch wavelet-transform dwt dtcwt scatternet filterbank differentiable gpu

Member repositories

For agents

markdown · JSON · MCP: product_card(name="fbcotter/pytorch_wavelets")