# sniklaus/sepconv-slomo

an implementation of Video Frame Interpolation via Adaptive Separable Convolution using PyTorch

Repository: https://github.com/sniklaus/sepconv-slomo
Canonical: https://ross.abutalabs.com/products/sepconv-slomo
Language: Python
License Family: other
Topics: pytorch, python, cuda, deep-learning, cupy
Last push: 2025-05-26T23:43:33+00:00

## Health v2 (maintenance only)
Score: 43/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 23, release rhythm 35, longevity 100
- inputs: {"age_days": 3279, "days_push": 464, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases, no_license
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1021, forks 166 (observed 2026-08-28T04:03:15.703405+00:00)

## What it is
A reference PyTorch implementation of Video Frame Interpolation via Adaptive Separable Convolution, which generates intermediate frames between two input frames or interpolates whole videos. It uses a custom CUDA separable convolution layer built with CuPy and is strictly for academic use.

## Use cases
- interpolate a middle frame between two images
- convert a video to slow motion by adding intermediate frames
- increase video frame rate with deep learning
- reproduce results from the ICCV 2017 separable convolution paper
- benchmark frame interpolation on Middlebury optical flow examples

## When to choose
- you need a faithful reference implementation of the adaptive separable convolution paper for research
- you want to interpolate frames or videos on a CUDA GPU with PyTorch
- you are doing academic work and can comply with the non-commercial license

## When to avoid
- you need commercial use - the license is academic-only
- you want the author's newer, improved model - use sniklaus/revisiting-sepconv instead
- you lack an NVIDIA GPU or cannot install CuPy

## Facets
- artifact type: library
- maturity: maintenance
- function: video-processing, deep-learning, image-processing
- domain: computer-vision, deep-learning, machine-learning
- platform: python
- tags: frame-interpolation, video-slow-motion, pytorch, cupy, optical-flow, academic-research, superseded, video, gpu, linux, cuda

## Member repositories
- sniklaus/sepconv-slomo (main) score 43

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:03:15.703405+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T07:08:52.074595+00:00, confidence not recorded.
  - readme: https://github.com/sniklaus/sepconv-slomo (fetched 2026-08-28T04:03:15.703405+00:00, sha e68120c3c772)
- Data as of 2026-08-30T08:39:29.467469+00:00.