Skip to content

ModelCard.md

Latest commit

 

History

History
84 lines (75 loc) · 7.49 KB

ModelCard.md

File metadata and controls

84 lines (75 loc) · 7.49 KB

Model Cards for IQA-PyTorch

General FR/NR Methods

List all model names with:

import pyiqa
print(pyiqa.list_models())
FR Method Model names Description
TOPIQ topiq_fr, topiq_fr-pipal Proposed in this paper
AHIQ ahiq
PieAPP pieapp
LPIPS lpips, lpips-vgg, stlpips, stlpips-vgg
DISTS dists
WaDIQaM No pretrain models
CKDN1 ckdn
FSIM fsim
SSIM ssim, ssimc Gray input (y channel), color input
MS-SSIM ms_ssim
CW-SSIM cw_ssim
PSNR psnr, psnry Color input, gray input (y channel)
VIF vif
GMSD gmsd
NLPD nlpd
VSI vsi
MAD mad
NR Method Model names Description
TOPIQ topiq_nr, topiq_nr-flive, topiq_nr-spaq TOPIQ with different datasets, koniq by default
TReS tres, tres-koniq, tres-flive TReS with different datasets, koniq by default
FID fid Statistic distance between two datasets
CLIPIQA(+) clipiqa, clipiqa+, clipiqa+_vitL14_512,clipiqa+_rn50_512 CLIPIQA(+) with different backbone, RN50 by default
MANIQA maniqa, maniqa-kadid, maniqa-koniq, maniqa-pipal MUSIQ with different datasets, koniq by default
MUSIQ musiq, musiq-koniq, musiq-spaq, musiq-paq2piq, musiq-ava MUSIQ with different datasets, koniq by default
DBCNN dbcnn
PaQ-2-PiQ paq2piq
HyperIQA hyperiqa
NIMA nima, nima-vgg16-ava Aesthetic metric trained with AVA dataset
WaDIQaM No pretrain models
CNNIQA cnniqa
NRQM(Ma)2 nrqm No backward
PI(Perceptual Index) pi No backward
BRISQUE brisque No backward
ILNIQE ilniqe No backward
NIQE niqe No backward

[1] This method use distorted image as reference. Please refer to the paper for details.
[2] Currently, only naive random forest regression is implemented and does not support backward.

IQA Methods for Specific Tasks

Task Method Description
Face IQA topiq_nr-face TOPIQ model trained with face IQA dataset (GFIQA)
Underwater IQA uranker A ranking-based underwater image quality assessment (UIQA) method, AAAI2023, Arxiv, Github

Outputs of Different Metrics

Note: ~ means that the corresponding numeric bound is typical value and not mathematically guaranteed

model lower better ? min max DATE Link
clipiqa False 0 1 2022 https://arxiv.org/abs/2207.12396
maniqa False 0 2022 https://arxiv.org/abs/2204.08958
hyperiqa False 0 1 2020 pdf
cnniqa False 2014 pdf
tres False 2022 https://github.com/isalirezag/TReS
musiq False ~0 ~100 2021 https://arxiv.org/abs/2108.05997
musiq-ava False ~0 ~10 2021 https://arxiv.org/abs/2108.05997
musiq-koniq False ~0 ~100 2021 https://arxiv.org/abs/2108.05997
musiq False 2021 https://arxiv.org/abs/2108.05997
paq2piq False 2020 pdf
dbcnn False 2019 https://arxiv.org/bas/1907.02665
brisque True 2012 pdf
pi True 2018 https://arxiv.org/abs/1809.07517
nima False 2018 https://arxiv.org/abs/1709.05424
nrqm False 2016 https://arxiv.org/abs/1612.05890
ilniqe True 0 2015 pdf
niqe True 0 2012 pdf