New advances in the detection of bias in face recognition algorithms
2021-01-25
(Press-News.org) A team from the Computer Vision Center (CVC) and the University of Barcelona has published the results of a study that evaluates the accuracy and bias in gender and skin colour of automatic face recognition algorithms tested with real world data. Although the top solutions exceed the 99.9% of accuracy, researchers have detected some groups that show higher false positive or false negative rates.
Face recognition has been routinely used by both private and governmental organizations worldwide. Automatic face recognition can be used for legitimate and beneficial purposes (e.g. to improve security) but at the same time its power and ubiquity increases a potential negative impact that unfair methods can have on society (e.g. discrimination against ethnic minorities). A necessary, albeit not sufficient, condition for a legitimate deployment of face recognition algorithms is the equal accuracy for all demographic groups.
With this purpose in mind, researchers from the Human Pose Recovery and Behavior Analysis Group at the Computer Vision Center (CVC) - University of Barcelona (UB), led by Sergio Escalera, organized a challenge within the European Conference of Computer Vision (ECCV) 2020. The results, recently published in the journal Computer Vision - ECCV 2020 Workshops, evaluated the accuracy of the submitted algorithms by the participants on the face verification task in the presence of other confounding attributes.
The challenge was a success, since "it attracted 151 participants, who made more than 1,800 submissions in total, exceeding our expectations regarding the number of participants and submissions" explained Sergio Escalera, also member of the Institute of Mathematics of the UB.
The participants used a not balanced image dataset, which simulates a real world scenario where AI based models are supposed to be trained and evaluated on imbalanced data (considerably more white males than dark females). In total, they worked with 152,917 images from 6,139 identities.
The images were annotated for two protected attributes: gender and skin colour; and five legitimate attributes: age group (0-34, 35-64, 65+), head pose (frontal, other), image source (still image, video frame), wearing glasses and a bounding box size.
The obtained results were very promising. Top winning solutions exceeded 99.9% of accuracy while achieving very low scores in the proposed bias metrics, "which can be considered a step toward the development of fairer face recognition methods" expounded Julio C. S. Jacques Jr., researcher at the CVC and at the Open University of Catalonia. The analysis of top 10 teams showed higher false positive rates for females with dark skin tone and for samples where both individuals wear glasses. In contrast, there were higher false negative rates for males with light skin tone and for samples where both individuals were aged 35 and below. Also, it was found that in the dataset individuals younger than 35 wear glasses less often than older individuals, resulting in a combination of effects of these attributes.
"This was not a surprise, since the adopted dataset was not balanced regarding the different demographic attributes. However, it shows that overall accuracy is not enough when the goal is to build fair face recognition methods, and that future works on the topic must take into account accuracy and bias mitigation together", concluded Julio C. S. Jacques Jr.
INFORMATION:
ELSE PRESS RELEASES FROM THIS DATE:
2021-01-25
Small-scale optical devices capable of using photons for high-speed information processing can be fabricated with unprecedented ease and precision using an additive manufacturing process developed at KAUST.
Fiber optics are conventionally produced by drawing thin filaments out of molten silica glass down to microscale dimensions. By infusing these fibers with long narrow hollow channels, a new class of optical devices termed "photonic crystal fibers" were introduced. The periodic arrangement of air holes in these photonic crystal fibers act like near-perfect mirrors, allowing trapping and long propagation of light in their central core.
"Photonic crystal fibers allow you to confine light in very tight spaces, increasing the optical ...
2021-01-25
Skeletal muscle is the largest tissue in the human body and is responsible not only for locomotion but for energy metabolism and heat production. Age-related muscle atrophy reduces motor function and contributes to the need for nursing care. In addition, muscle atrophy associated with various chronic diseases is known to be a risk factor for life expectancy. Myoblasts, the progenitor cells of muscle, play an important role in maintaining muscle homeostasis. However, it has been reported that the differentiation ability of myoblasts decreases with age and disease, and this is thought to be one of the causes of muscle atrophy. In order to prevent or treat muscle atrophy, a research team led by Assistant Professor Tomohide Takaya of Shinshu University, Faculty of Agriculture studied ...
2021-01-25
Laser beams can be used to precisely measure an object's position or velocity. Normally, however, a clear, unobstructed view of this object is required - and this prerequisite is not always satisfied. In biomedicine, for example, structures are examined, which are embedded in an irregular, complicated environment. There, the laser beam is deflected, scattered and refracted, often making it impossible to obtain useful data from the measurement.
However, Utrecht University (Netherlands) and TU Wien (Vienna, Austria) have now been able to show that meaningful results can be obtained even in such complicated environments. Indeed, there is a way to specifically modify ...
2021-01-25
Depression, especially in urban areas, is on the rise, now more than ever. Mental health outcomes are influenced by, among other things, the type of environment where one lives. Former studies show that urban greenspace has a positive benefit on people experiencing mental ill health, but most of these studies used self-reported measures, which makes it difficult to compare the results and generalise conclusions on the effects of urban greenspace on mental health.
An interdisciplinary research team of UFZ, iDiv and Leipzig University tried to improve this issue by involving an objective indicator: prescriptions of antidepressants. To find out whether a specific type of 'everyday' green space - street trees dotting the neighbourhood sidewalks ...
2021-01-25
A lot is known about galaxies. We know, for instance, that the stars within them are shaped from a blend of old star dust and molecules suspended in gas. What remains a mystery, however, is the process that leads to these simple elements being pulled together to form a new star.
But now an international team of scientists, including astrophysicists from the University of Bath in the UK and the National Astronomical Observatory (OAN) in Madrid, Spain have taken a significant step towards understanding how a galaxy's gaseous content becomes organised into a new generation of stars.
Their findings have important implications for our understanding of how stars formed during the early days of the universe, when galaxy collisions were frequent and dramatic, and star and galaxy formation ...
2021-01-25
Children of mothers experiencing depressive symptoms are more at risk, as adolescents, of experiencing suicidal thoughts and attempting suicide.
New research suggests that this link may be explained by loneliness, potentially opening new ways for youth suicide prevention.
The study - by the universities of Exeter, Montréal, Laval and McGill - used data from more than 1,600 families from the Québec Longitudinal Study of Child Development, a representative sample of new-borns in Quebec followed from birth to 20 years of age.
Mothers were asked about depressive symptoms (such as sadness and losing interest ...
2021-01-25
As the world becomes aware of the imminent environmental crisis, scientists have begun a search for sustainable energy sources. Rechargeable batteries like lithium-ion batteries are seeing a popularity surge, concurrent with production of "greener" technologies such as electric propulsion ships (which are being developed to meet the environmental regulations by the International Maritime Organization) and other electric vehicles. But, lithium is rare and difficult to distribute, putting its sustainability in doubt while also risking sharp increases in cost. Researchers have thus turned ...
2021-01-25
To predict when earthquakes are likely to occur, seismologists often use statistics to monitor how clusters of seismic activity evolve over time. However, this approach often fails to anticipate the time and magnitude of large-scale earthquakes, leading to dangerous oversights in current early-warning systems. For decades, studies outside the seismology field have proposed that these major, potentially devastating seismic events are connected to a range of non-seismic phenomena - which can be observed days or even weeks before these large earthquakes occur. So far, however, this idea hasn't caught on in the wider scientific community. In this special issue, EPJ Special Topics proposes the Global Earthquake Forecasting System (GEFS): the first collaborative initiative ...
2021-01-25
Scientists have shown where bird species would exist in the absence of human activity under research that could provide a new approach to setting conservation priorities.
A study by Durham University, UK, in collaboration with the Royal Society for the Protection of Birds (RSPB), investigated how human activities such as agriculture, deforestation, and the drainage of wetlands have shaped where bird species are found in Great Britain today.
Researchers used data on the geographical distributions of bird species alongside simulation models to predict where bird species would exist today if the effects of human activities on the landscape were removed.
In this scenario there were winners ...
2021-01-25
Public health researchers, led by UNSW Sydney, have estimated the number of cancer cases requiring surgery globally each year, predicting the number will rise from 9.1 million to 13.8 million from 2018 to 2040 - an increase of 52 per cent or 4.7 million cases.
Their research shows the greatest relative increase will occur in 34 low-income countries, where the number of cases requiring surgery is expected to more than double by 2040 (314,355 cases to 650,164, or 107 per cent).
The modelling study, published in The Lancet Oncology on Friday, analysed global demand for cancer surgery and estimated surgical and anaesthesia ...
LAST 30 PRESS RELEASES:
[Press-News.org] New advances in the detection of bias in face recognition algorithms
