# IQA-PyTorch

🔎 🖼️ 🔥PyTorch Toolbox for Image Quality Assessment, including PSNR, SSIM, LPIPS, FID, NIQE, NRQM(Ma), MUSIQ, TOPIQ, NIMA, DBCNN, BRISQUE, PI and more...

Repository: https://github.com/chaofengc/IQA-PyTorch
Canonical: https://ross.abutalabs.com/products/iqa-pytorch
Homepage: https://iqa-pytorch.readthedocs.io/
Language: Python
License: NOASSERTION
License Family: other
Topics: iqa, image-quality-assessment, blind-image-quality-assessment, pytorch, pytorch-implementation, python, image-aesthetic-assessment, ms-ssim, psnr, ssim
Last push: 2026-07-08T10:48:34+00:00

## Health v2 (maintenance only)
Score: 82/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 91, release rhythm 60, longevity 100
- inputs: {"age_days": 1739, "days_push": 56, "days_rel": 56, "gap_med": 246, "n_releases_24m": 4}
- flags: no_license
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 3380, forks 252 (observed 2026-08-28T04:07:58.566303+00:00)

## What it is
A pure Python/PyTorch toolbox for image quality assessment (IQA) providing GPU-accelerated reimplementations of many full-reference and no-reference metrics such as PSNR, SSIM, LPIPS, FID, NIQE, MUSIQ, TOPIQ, and NIMA. Results are calibrated against official MATLAB implementations, and it is installable via pip as `pyiqa`.

## Use cases
- compute image quality metrics like psnr ssim lpips in python
- score images with no-reference quality models like musiq or nima
- evaluate generated images with fid
- benchmark image restoration models against quality metrics
- replace slow matlab iqa implementations with fast gpu pytorch versions
- assess aesthetic quality of photos
- measure perceptual image quality for dataset curation

## When to choose
- you need many IQA metrics behind one consistent Python API
- you want GPU-accelerated metric computation in a PyTorch pipeline
- you need pretrained no-reference models like MUSIQ, TOPIQ, or NIMA without reimplementing them
- you want results calibrated against official MATLAB reference implementations

## When to avoid
- you need a single trivial metric and want zero deep-learning dependencies
- you require strict license compliance and cannot accept a non-standard license
- you work outside Python/PyTorch ecosystems, e.g. pure MATLAB or TensorFlow

## Facets
- artifact type: library
- maturity: active
- function: image-processing, machine-learning, benchmarking, sdk
- domain: computer-vision, image-processing, machine-learning, deep-learning
- platform: python, cross-platform
- tags: image-quality-assessment, pytorch, iqa-metrics, no-reference-metrics, full-reference-metrics, psnr, ssim, lpips, fid, musiq, nima, aesthetic-assessment, pip-installable, gpu

## Member repositories
- chaofengc/IQA-PyTorch (main) score 82
- chaofengc/Awesome-Image-Quality-Assessment (docs) score 75

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:07:58.566303+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 2026-08-29T18:39:41.160222+00:00, confidence not recorded.
  - readme: https://github.com/chaofengc/IQA-PyTorch (fetched 2026-08-28T04:07:58.566303+00:00, sha 0280d90739be)
- Data as of 2026-08-30T08:39:29.467469+00:00.