Maghoumi/pytorch-softdtw-cuda

Fast CUDA implementation of (differentiable) soft dynamic time warping for PyTorch

44
/ 100
Emerging

This helps deep learning researchers and practitioners using PyTorch to efficiently compare time series data, particularly for tasks like gesture recognition or motion analysis. It takes two batches of time series sequences as input and outputs a differentiable similarity score, much like a loss function. This tool is for machine learning engineers and researchers working with sequential data who need faster computation of dynamic time warping.

728 stars. No commits in the last 6 months.

Use this if you are working with time series data in PyTorch and need to compute Soft Dynamic Time Warping distances much faster than CPU-based implementations, especially for larger datasets and longer sequences.

Not ideal if your input sequences are extremely long (over 1024 data points) or if you are not using PyTorch and CUDA for your deep learning models.

time-series-analysis gesture-recognition deep-learning-research motion-analysis sequence-comparison
Stale 6m No Package No Dependents
Maintenance 0 / 25
Adoption 10 / 25
Maturity 16 / 25
Community 18 / 25

How are scores calculated?

Stars

728

Forks

66

Language

Python

License

MIT

Last pushed

Apr 03, 2024

Commits (30d)

0

Get this data via API

curl "https://pt-edge.onrender.com/api/v1/quality/ml-frameworks/Maghoumi/pytorch-softdtw-cuda"

Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.