2021
DOI: 10.3390/math9020195
|View full text |Cite
|
Sign up to set email alerts
|

Recognizing Human Races through Machine Learning—A Multi-Network, Multi-Features Study

Abstract: The human face holds a privileged position in multi-disciplinary research as it conveys much information—demographical attributes (age, race, gender, ethnicity), social signals, emotion expression, and so forth. Studies have shown that due to the distribution of ethnicity/race in training datasets, biometric algorithms suffer from “cross race effect”—their performance is better on subjects closer to the “country of origin” of the algorithm. The contributions of this paper are two-fold: (a) first, we gathered, … Show more

Help me understand this report

Search citation statements

Order By: Relevance

Paper Sections

Select...
2
2
1

Citation Types

0
14
0

Year Published

2022
2022
2024
2024

Publication Types

Select...
4
4

Relationship

0
8

Authors

Journals

citations
Cited by 17 publications
(14 citation statements)
references
References 47 publications
0
14
0
Order By: Relevance
“…We conducted testing of the trained CNN using two publicly available datasets: UTKFace 62 and FaceARG 63 . The UTKFace dataset consists of over 20,000 face images with dimensions of .…”
Section: Methodsmentioning
confidence: 99%
“…We conducted testing of the trained CNN using two publicly available datasets: UTKFace 62 and FaceARG 63 . The UTKFace dataset consists of over 20,000 face images with dimensions of .…”
Section: Methodsmentioning
confidence: 99%
“…This subsection is devoted to introducing a performance comparison considering several contributions from the SoTA. In recent years, numerous additional studies have attempted to address the classification problems of race [4] and ethnicity [46] by means of DL approaches, e.g., CNNs. Table 1 presents the results provided by several methods of the SoTA dealing with the tackled problem.…”
Section: A Preliminary Comparitive Analysismentioning
confidence: 99%
“…Note that it was not feasible to carry out a direct numerical comparison against all methods due to several of them were trained and assessed on not accessible private datasets. In [4], the authors developed and made available a large-scale library of over 175,000 photos of faces of celebrities from multiple ethnicities and races, which were used for training and testing. The method was compared against four cutting-edge CNNs on the topic.…”
Section: A Preliminary Comparitive Analysismentioning
confidence: 99%
See 1 more Smart Citation
“…In an attempt to quantify and mitigate varying accuracy across race in facial recognition systems, many researchers have encouraged the use of racially balanced datasets over the use of those that are racially unbalanced [10][11][12][13]. When a racially balanced dataset is not available for training, a racially balanced evaluation is encouraged [14][15][16][17].…”
Section: Introductionmentioning
confidence: 99%