-
Can the accuracy bias by facial hairstyle be reduced through balancing the training data?
Authors:
Kagan Ozturk,
Haiyu Wu,
Kevin W. Bowyer
Abstract:
Appearance of a face can be greatly altered by growing a beard and mustache. The facial hairstyles in a pair of images can cause marked changes to the impostor distribution and the genuine distribution. Also, different distributions of facial hairstyle across demographics could cause a false impression of relative accuracy across demographics. We first show that, even though larger training sets b…
▽ More
Appearance of a face can be greatly altered by growing a beard and mustache. The facial hairstyles in a pair of images can cause marked changes to the impostor distribution and the genuine distribution. Also, different distributions of facial hairstyle across demographics could cause a false impression of relative accuracy across demographics. We first show that, even though larger training sets boost the recognition accuracy on all facial hairstyles, accuracy variations caused by facial hairstyles persist regardless of the size of the training set. Then, we analyze the impact of having different fractions of the training data represent facial hairstyles. We created balanced training sets using a set of identities available in Webface42M that both have clean-shaven and facial hair images. We find that, even when a face recognition model is trained with a balanced clean-shaven / facial hair training set, accuracy variation on the test data does not diminish. Next, data augmentation is employed to further investigate the effect of facial hair distribution in training data by manipulating facial hair pixels with the help of facial landmark points and a facial hair segmentation model. Our results show facial hair causes an accuracy gap between clean-shaven and facial hair images, and this impact can be significantly different between African-Americans and Caucasians.
△ Less
Submitted 30 May, 2024;
originally announced May 2024.
-
Identity Overlap Between Face Recognition Train/Test Data: Causing Optimistic Bias in Accuracy Measurement
Authors:
Haiyu Wu,
Sicong Tian,
Jacob Gutierrez,
Aman Bhatta,
Kağan Öztürk,
Kevin W. Bowyer
Abstract:
A fundamental tenet of pattern recognition is that overlap between training and testing sets causes an optimistic accuracy estimate. Deep CNNs for face recognition are trained for N-way classification of the identities in the training set. Accuracy is commonly estimated as average 10-fold classification accuracy on image pairs from test sets such as LFW, CALFW, CPLFW, CFP-FP and AgeDB-30. Because…
▽ More
A fundamental tenet of pattern recognition is that overlap between training and testing sets causes an optimistic accuracy estimate. Deep CNNs for face recognition are trained for N-way classification of the identities in the training set. Accuracy is commonly estimated as average 10-fold classification accuracy on image pairs from test sets such as LFW, CALFW, CPLFW, CFP-FP and AgeDB-30. Because train and test sets have been independently assembled, images and identities in any given test set may also be present in any given training set. In particular, our experiments reveal a surprising degree of identity and image overlap between the LFW family of test sets and the MS1MV2 training set. Our experiments also reveal identity label noise in MS1MV2. We compare accuracy achieved with same-size MS1MV2 subsets that are identity-disjoint and not identity-disjoint with LFW, to reveal the size of the optimistic bias. Using more challenging test sets from the LFW family, we find that the size of the optimistic bias is larger for more challenging test sets. Our results highlight the lack of and the need for identity-disjoint train and test methodology in face recognition research.
△ Less
Submitted 15 May, 2024;
originally announced May 2024.
-
Hyperfine Structure Investigation of Singly Ionized Thulium in FT Spectra
Authors:
T. Y. Kebapci,
S. Parlatan,
S. Sert,
I. K. Ozturk,
G. Basar,
T. Sahin,
S. Bilir,
R. Ferber,
M. Tamanis,
S. Kroger
Abstract:
The hyperfine structure of 40 spectral lines of singly ionized thulium (Tm II) in emission spectra from a hollow cathode discharge lamp measured with a Fourier transform spectrometer in the wavelength range from 335 nm to 2345 nm has been investigated. As a result of the analysis, the magnetic dipole hyperfine structure constants $A$ for 27 fine structure levels of Tm II were determined for the fi…
▽ More
The hyperfine structure of 40 spectral lines of singly ionized thulium (Tm II) in emission spectra from a hollow cathode discharge lamp measured with a Fourier transform spectrometer in the wavelength range from 335 nm to 2345 nm has been investigated. As a result of the analysis, the magnetic dipole hyperfine structure constants $A$ for 27 fine structure levels of Tm II were determined for the first time. In addition, the values of two magnetic dipole hyperfine structure constants $A$ from the literature were declared incorrect and the corrected values were given.
△ Less
Submitted 8 May, 2024;
originally announced May 2024.
-
Beard Segmentation and Recognition Bias
Authors:
Kagan Ozturk,
Grace Bezold,
Aman Bhatta,
Haiyu Wu,
Kevin Bowyer
Abstract:
A person's facial hairstyle, such as presence and size of beard, can significantly impact face recognition accuracy. There are publicly-available deep networks that achieve reasonable accuracy at binary attribute classification, such as beard / no beard, but few if any that segment the facial hair region. To investigate the effect of facial hair in a rigorous manner, we first created a set of fine…
▽ More
A person's facial hairstyle, such as presence and size of beard, can significantly impact face recognition accuracy. There are publicly-available deep networks that achieve reasonable accuracy at binary attribute classification, such as beard / no beard, but few if any that segment the facial hair region. To investigate the effect of facial hair in a rigorous manner, we first created a set of fine-grained facial hair annotations to train a segmentation model and evaluate its accuracy across African-American and Caucasian face images. We then use our facial hair segmentations to categorize image pairs according to the degree of difference or similarity in the facial hairstyle. We find that the False Match Rate (FMR) for image pairs with different categories of facial hairstyle varies by a factor of over 10 for African-American males and over 25 for Caucasian males. To reduce the bias across image pairs with different facial hairstyles, we propose a scheme for adaptive thresholding based on facial hairstyle similarity. Evaluation on a subject-disjoint set of images shows that adaptive similarity thresholding based on facial hairstyles of the image pair reduces the ratio between the highest and lowest FMR across facial hairstyle categories for African-American from 10.7 to 1.8 and for Caucasians from 25.9 to 1.3. Facial hair annotations and facial hair segmentation model will be publicly available.
△ Less
Submitted 29 August, 2023;
originally announced August 2023.
-
Experimental Investigation of the Hyperfine Structure of Tm I with Fourier Transform Spectroscopy Part B: in the NIR wavelength range from 700 nm to 2250 nm
Authors:
Taha Yusuf Kebapcı,
Sami Sert,
Şeyma Parlatan,
İpek Kanat Öztürk,
Gönül Başar,
Günay Başar,
Maris Tamanis,
Sophie Kröger
Abstract:
In this study, we investigated the hyperfine structure of 43 spectral lines of atomic thulium. We analyzed Fourier-transform spectra in the wavelength range from 700 nm to 2250 nm, which corresponds to the wavenumber range from 14300 cm-1 to 4440 cm-1, respectively. The excited thulium atoms were generated in a hollow-cathode lamp. As a result of this investigation, the magnetic-dipole hyperfine c…
▽ More
In this study, we investigated the hyperfine structure of 43 spectral lines of atomic thulium. We analyzed Fourier-transform spectra in the wavelength range from 700 nm to 2250 nm, which corresponds to the wavenumber range from 14300 cm-1 to 4440 cm-1, respectively. The excited thulium atoms were generated in a hollow-cathode lamp. As a result of this investigation, the magnetic-dipole hyperfine constant A of 17 fine structure levels have been determined experimentally, 14 of them for the first time. The magnetic-dipole hyperfine constant values of the three remaining levels, reported in the literature, differed significantly from the results of our determination.
△ Less
Submitted 5 April, 2022;
originally announced April 2022.
-
Experimental Investigation of the Hyperfine Structure of Tm I with Fourier Transform Spectroscopy, Part A: in the visible wavelength range 400 nm 700 nm
Authors:
Seyma Parlatan,
Ipek Kanat Ozturk,
Gonul Basar,
Gunay Basar,
Ruvin Ferber,
Sophie Kroger
Abstract:
The spectra emitted by a thulium hollow-cathode discharge lamp with argon or neon as inert gases have been recorded with a Bruker IFS 125 HR Fourier Transform spectrometer in the visible wavelength region from 400 nm to 700 nm. The paper presents the investigation of the hyperfine-structure splitting of 50 spectral lines of neutral thulium. As a result, the magnetic dipole hyperfine structure cons…
▽ More
The spectra emitted by a thulium hollow-cathode discharge lamp with argon or neon as inert gases have been recorded with a Bruker IFS 125 HR Fourier Transform spectrometer in the visible wavelength region from 400 nm to 700 nm. The paper presents the investigation of the hyperfine-structure splitting of 50 spectral lines of neutral thulium. As a result, the magnetic dipole hyperfine structure constant A for 20 fine structure levels of Tm I have been determined for the first time.
△ Less
Submitted 5 April, 2022;
originally announced April 2022.
-
The RSNA-ASNR-MICCAI BraTS 2021 Benchmark on Brain Tumor Segmentation and Radiogenomic Classification
Authors:
Ujjwal Baid,
Satyam Ghodasara,
Suyash Mohan,
Michel Bilello,
Evan Calabrese,
Errol Colak,
Keyvan Farahani,
Jayashree Kalpathy-Cramer,
Felipe C. Kitamura,
Sarthak Pati,
Luciano M. Prevedello,
Jeffrey D. Rudie,
Chiharu Sako,
Russell T. Shinohara,
Timothy Bergquist,
Rong Chai,
James Eddy,
Julia Elliott,
Walter Reade,
Thomas Schaffter,
Thomas Yu,
Jiaxin Zheng,
Ahmed W. Moawad,
Luiz Otavio Coelho,
Olivia McDonnell
, et al. (78 additional authors not shown)
Abstract:
The BraTS 2021 challenge celebrates its 10th anniversary and is jointly organized by the Radiological Society of North America (RSNA), the American Society of Neuroradiology (ASNR), and the Medical Image Computing and Computer Assisted Interventions (MICCAI) society. Since its inception, BraTS has been focusing on being a common benchmarking venue for brain glioma segmentation algorithms, with wel…
▽ More
The BraTS 2021 challenge celebrates its 10th anniversary and is jointly organized by the Radiological Society of North America (RSNA), the American Society of Neuroradiology (ASNR), and the Medical Image Computing and Computer Assisted Interventions (MICCAI) society. Since its inception, BraTS has been focusing on being a common benchmarking venue for brain glioma segmentation algorithms, with well-curated multi-institutional multi-parametric magnetic resonance imaging (mpMRI) data. Gliomas are the most common primary malignancies of the central nervous system, with varying degrees of aggressiveness and prognosis. The RSNA-ASNR-MICCAI BraTS 2021 challenge targets the evaluation of computational algorithms assessing the same tumor compartmentalization, as well as the underlying tumor's molecular characterization, in pre-operative baseline mpMRI data from 2,040 patients. Specifically, the two tasks that BraTS 2021 focuses on are: a) the segmentation of the histologically distinct brain tumor sub-regions, and b) the classification of the tumor's O[6]-methylguanine-DNA methyltransferase (MGMT) promoter methylation status. The performance evaluation of all participating algorithms in BraTS 2021 will be conducted through the Sage Bionetworks Synapse platform (Task 1) and Kaggle (Task 2), concluding in distributing to the top ranked participants monetary awards of $60,000 collectively.
△ Less
Submitted 12 September, 2021; v1 submitted 5 July, 2021;
originally announced July 2021.
-
HAMSI: A Parallel Incremental Optimization Algorithm Using Quadratic Approximations for Solving Partially Separable Problems
Authors:
Kamer Kaya,
Figen Öztoprak,
Ş. İlker Birbil,
A. Taylan Cemgil,
Umut Şimşekli,
Nurdan Kuru,
Hazal Koptagel,
M. Kaan Öztürk
Abstract:
We propose HAMSI (Hessian Approximated Multiple Subsets Iteration), which is a provably convergent, second order incremental algorithm for solving large-scale partially separable optimization problems. The algorithm is based on a local quadratic approximation, and hence, allows incorporating curvature information to speed-up the convergence. HAMSI is inherently parallel and it scales nicely with t…
▽ More
We propose HAMSI (Hessian Approximated Multiple Subsets Iteration), which is a provably convergent, second order incremental algorithm for solving large-scale partially separable optimization problems. The algorithm is based on a local quadratic approximation, and hence, allows incorporating curvature information to speed-up the convergence. HAMSI is inherently parallel and it scales nicely with the number of processors. Combined with techniques for effectively utilizing modern parallel computer architectures, we illustrate that the proposed method converges more rapidly than a parallel stochastic gradient descent when both methods are used to solve large-scale matrix factorization problems. This performance gain comes only at the expense of using memory that scales linearly with the total size of the optimization variables. We conclude that HAMSI may be considered as a viable alternative in many large scale problems, where first order methods based on variants of stochastic gradient descent are applicable.
△ Less
Submitted 4 August, 2017; v1 submitted 5 September, 2015;
originally announced September 2015.
-
Dynamics of discrete opinions without compromise
Authors:
M. Kaan Öztürk
Abstract:
A new agent-based, bounded-confidence model for discrete one-dimensional opinion dynamics is presented. The agents interact if their opinions do not differ more than a tolerance parameter. In pairwise interactions, one of the pair, randomly selected, converts to the opinion of the other. The model can be used to simulate cases where no compromise is possible, such as choices of substitute goods, o…
▽ More
A new agent-based, bounded-confidence model for discrete one-dimensional opinion dynamics is presented. The agents interact if their opinions do not differ more than a tolerance parameter. In pairwise interactions, one of the pair, randomly selected, converts to the opinion of the other. The model can be used to simulate cases where no compromise is possible, such as choices of substitute goods, or other exclusive choices. The homogeneous case with maximum tolerance is equivalent to the Gambler's Ruin problem. A homogeneous system always ends up in an absorbing state, which can have one or more surviving opinions. An upper bound for the final number of opinions is given. The distribution of absorption times fits the generalized extreme value distribution. The diffusion coefficient of an opinion increases linearly with the number of opinions within the tolerance parameter. A general master equation and specific Markov matrices are given. The software code developed for this study is provided as a supplement.
△ Less
Submitted 17 February, 2013; v1 submitted 20 December, 2011;
originally announced December 2011.
-
Trajectories of charged particles trapped in Earth's magnetic field
Authors:
M. Kaan Öztürk
Abstract:
I outline the theory of relativistic charged-particle motion in the magnetosphere in a way suitable for undergraduate courses. I discuss particle and guiding center motion, derive the three adiabatic invariants associated with them, and present particle trajectories in a dipolar field. I provide twelve computational exercises that can be used as classroom assignments or for self-study. Two of the…
▽ More
I outline the theory of relativistic charged-particle motion in the magnetosphere in a way suitable for undergraduate courses. I discuss particle and guiding center motion, derive the three adiabatic invariants associated with them, and present particle trajectories in a dipolar field. I provide twelve computational exercises that can be used as classroom assignments or for self-study. Two of the exercises, drift-shell bifurcation and Speiser orbits, are adapted from active magnetospheric research. The Python code provided in the supplement can be used to replicate the trajectories and can be easily extended for different field geometries.
△ Less
Submitted 15 December, 2011;
originally announced December 2011.
-
Calculation of the elastic scattering parameters in an ultra-cold Fermi-Bose and Bose-Bose Fr vapor
Authors:
M. Kemal Öztürk,
Süleyman Özçelik
Abstract:
The calculations of the elastic scattering properties of ultracold-francium isotopes are reported in the detail. A parametric model potential for the triplet molecular states of the Fr$_{2}$ is represented, and the scattering lengths at and the effective range $r_{e}$ are calculated using WKB and Numerov methods for Fr-Fr in the triplet state. The convergence of these scattering properties as th…
▽ More
The calculations of the elastic scattering properties of ultracold-francium isotopes are reported in the detail. A parametric model potential for the triplet molecular states of the Fr$_{2}$ is represented, and the scattering lengths at and the effective range $r_{e}$ are calculated using WKB and Numerov methods for Fr-Fr in the triplet state. The convergence of these scattering properties as the depending on a $K_{0}$ parameter and core radius is also investigated using two method as Quantum Defect Theory and analytical formula of symskovisky.
△ Less
Submitted 7 June, 2004;
originally announced June 2004.
-
Calculation of the elastic scattering properties in an ultra-cold Fermi-Bose and Bose-Bose Rb-K vapor
Authors:
M. Kemal Öztürk,
Süleyman Özçelik
Abstract:
The calculation of the elastic scattering properties of mixtures composed by rubidium and potassium atoms is reported and compared with experimental results in detail. The improved potentials for both molecular states the singlet and the triplet of the RbK are presented, and the scattering lengths a_{t} and the effective range r_{e} are calculated with the help of this potential using WKB and Nu…
▽ More
The calculation of the elastic scattering properties of mixtures composed by rubidium and potassium atoms is reported and compared with experimental results in detail. The improved potentials for both molecular states the singlet and the triplet of the RbK are presented, and the scattering lengths a_{t} and the effective range r_{e} are calculated with the help of this potential using WKB and Numerov methods for Rb-K in the triplet and singlet state. In addition, the convergence of these scattering properties as the depending on a K_{0} parameter is investigated using Quantum Defect Theory. The evaporative cooling other results that include the cross section and the spin-charge cross section, the rate coefficient as a function of the energy and the ultra-low temperature are also presented in this study.
△ Less
Submitted 12 March, 2004;
originally announced March 2004.
