LossFIQA: A Shortcut Solution to Image Quality Assessment Using Loss for Faces and Beyond
The result's identifiers
Result code in IS VaVaI
<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F00216305%3A26230%2F26%3A0198680" target="_blank" >RIV/00216305:26230/26:0198680 - isvavai.cz</a>
Result on the web
<a href="https://ieeexplore.ieee.org/document/11082134" target="_blank" >https://ieeexplore.ieee.org/document/11082134</a>
DOI - Digital Object Identifier
<a href="http://dx.doi.org/10.1109/ACCESS.2025.3589778" target="_blank" >10.1109/ACCESS.2025.3589778</a>
Alternative languages
Result language
angličtina
Original language name
LossFIQA: A Shortcut Solution to Image Quality Assessment Using Loss for Faces and Beyond
Original language description
We introduce a novel approach to model-based quality assessment of input images. Our approach is very simple, and we demonstrate experimentally that it is not limited to a single domain (typically face recognition in the literature). Our approach generates per-sample quality pseudo-labels directly from the objective function used during the training of the target model. We evaluate the proposed method on eight large and respected datasets (from the face recognition on LFW, CALFW, CPLFW, XQLFW, CFP-FP, AgeDB, IJB-C, and retinopathy detection domain on EyePACS dataset) and using multiple state-of-the-art models (AdaFace, MagFace, ArcFace, ElasticFace, and CuricularFace). Compared to state-of-the-art methods for face quality assessment that are considerably more complex, our solution yields competitive results while being much simpler and not limited to one application.
Czech name
—
Czech description
—
Classification
Type
J<sub>imp</sub> - Article in a specialist periodical, which is included in the Web of Science database
CEP classification
—
OECD FORD branch
10201 - Computer sciences, information science, bioinformathics (hardware development to be 2.2, social aspect to be 5.8)
Result continuities
Project
—
Continuities
S - Specificky vyzkum na vysokych skolach
Others
Publication year
2025
Confidentiality
S - Úplné a pravdivé údaje o projektu nepodléhají ochraně podle zvláštních právních předpisů
Data specific for result type
Name of the periodical
IEEE Access
ISSN
2169-3536
e-ISSN
—
Volume of the periodical
13
Issue of the periodical within the volume
7
Country of publishing house
US - UNITED STATES
Number of pages
10
Pages from-to
126915-126924
UT code for WoS article
001536742300019
EID of the result in the Scopus database
2-s2.0-105011055485