Age Estimation Performance Research Articles

Accuracy of the London Atlas to estimate the age of legal majority in a sample of the Amazon Regio.

Estimating the age of majority is a challenging task in forensic odontology, especially because the third molars are usually the only developing teeth between the ages of 16 and 21 years. The London Atlas emerged as an alternative to estimate age using dental development, eruption and deciduous root resorption as parameters. The method has performed well in young age categories, while its performance for age estimation via third molars has been disputed. The present study tested the performance of the London Atlas to estimate the age of legal majority in a sample of individuals from the Amazon Region. The sample consisted of 1.256 panoramic radiographs of women (n = 694) and men (n = 562) between 16 and 22.9 years. The method was applied to the maxillary (#28) and mandibular (#38) left third molars. For comparative purposes, the sample was divided into seven age groups: 16├ 16.99; 17├ 17.99; 18├ 18.99; 19├ 19.99; 20├ 20.99; 21├ 21.99; and 22├ 22.99 years. Chronological and estimated ages were compared descriptively by means of mean absolute errors (MAE) and root mean squared errors (RMSE), as well as through Receiver Operating Characteristic (ROC) curves and their area under the curve (AUC). The MAE of the age estimates using tooth #28 was 1.76 years for females and 1.52 years for males. When the tooth #38 was used, the MAE for the females and males were 1.68 and 1.51 years, respectively. The MAE and RMSE increased in ascending order between age categories. Tooth #28 led to 74% of correct classifications around the age of legal majority, while tooth #38 reached 77%. The area under the curve was 0.75 for tooth #28 and 0.73 for tooth #38. The London Atlas should be used carefully to estimate the age of legal majority and not as a single method when the age threshold is 18 years. Key words:Age determination by teeth, forensic dentistry, forensic science, third molar.

The effect of reference sample composition and size on dental age interval estimates.

Validation studies in juvenile dental age estimation primarily focus on point estimates while interval performance for reference samples of different ancestry group compositions has received minimal attention. We tested the effect of reference sample size and composition by sex and ancestry group on age interval estimates. The dataset consisted of Moorrees et al. dental scores from panoramic radiographs of 3334 London children of Bangladeshi and European ancestry and 2-23 years of age. Model stability was assessed using standard error of mean age-at-transition for univariate cumulative probit and sample size, group mixing (sex or ancestry), and staging system as factors. Age estimation performance was tested using molar reference samples of four sizes, stratified by year of age, sex, and ancestry. Age estimates were performed using Bayesian multivariate cumulative probit with 5-fold cross-validation. Standard error increased with decreasing sample size but showed no effect from mixing by sex or ancestry. Estimating ages using a reference and target sample of different sex reduced success rate significantly. The same test by ancestry groups had a lesser effect. Small sample size (n < 20/year of age) negatively affected most performance metrics. We found that reference sample size, followed by sex, primarily drove age estimation performance. Combining reference samples by ancestry produced equivalent or better estimates of age by all metrics than using a single-demographic reference of smaller size. We further proposed that population specificity is an alternative hypothesis of intergroup difference that has been erroneously treated as a null.

Applicability of Shoulder, Olecranon, and Wrist-based Skeletal Maturity Estimation Systems to the Modern Pediatric Population.

The proximal humerus ossification system (PHOS), olecranon apophyseal ossification system (OAOS), and modified Fels wrist skeletal maturity system (mFWS) were recently developed or updated using a historical, mostly White, pediatric population. These upper extremity skeletal maturity systems have demonstrated skeletal age estimation performance superior or equivalent to Greulich and Pyle in historical patients. Their applicability to modern pediatric populations has not yet been evaluated. We reviewed anteroposterior shoulder, lateral elbow, and anteroposterior hand and wrist x-rays of 4 pediatric cohorts: White males, Black males, White females, and Black females. Peripubertal x-rays were evaluated: males 9 to17 years and females 7 to 15 years. Five nonpathologic radiographs for each age and joint were randomly selected from each group. Skeletal age estimates made by each of the 3 skeletal maturity systems were plotted against the chronological age associated with each radiograph and compared between cohorts, and with the historical patients. Five hundred forty modern radiographs were evaluated (180 shoulders, 180 elbows, and 180 wrists). All radiographic parameters had inter- and intra-rater reliability coefficients at or above 0.79, indicating very good reliability. For PHOS, White males had delayed skeletal age compared with Black males (Δ-0.12y, P =0.02) and historical males (Δ-0.17y, P <0.001). Black females were skeletally advanced compared with historical females (Δ0.11y, P =0.01). For OAOS, White males (Δ-0.31y, P <0.001) and Black males (Δ-0.24y, P <0.001) had delayed skeletal age compared with historical males. For mFWS, White males (Δ0.29y, P =0.024), Black males (Δ0.58y, P <0.001), and Black females (Δ0.44y, P <0.001) had advanced skeletal age compared with historical counterparts of the same sex. All other comparisons were not significant ( P >0.05). The PHOS, OAOS, and mFWS have mild discrepancies in skeletal age estimates when applied to modern pediatric populations depending on the race and sex of the patient. Level III - retrospective chart review.

High performance for bone age estimation with an artificial intelligence solution

PurposeThe purpose of this study was to compare the performance of an artificial intelligence (AI) solution to that of a senior general radiologist for bone age assessment. Material and methodsAnteroposterior hand radiographs of eight boys and eight girls from each age interval between five and 17 year-old from four different radiology departments were retrospectively collected. Two board-certified pediatric radiologists with knowledge of the sex and chronological age of the patients independently estimated the Greulich and Pyle bone age to determine the standard of reference. A senior general radiologist not specialized in pediatric radiology (further referred to as “the reader”) then determined the bone age with knowledge of the sex and chronological age. The results of the reader were then compared to those of the AI solution using mean absolute error (MAE) in age estimation. ResultsThe study dataset included a total of 206 patients (102 boys of mean chronological age of 10.9 ± 3.7 [SD] years, 104 girls of mean chronological age of 11 ± 3.7 [SD] years). For both sexes, the AI algorithm showed a significantly lower MAE than the reader (P < 0.007). In boys, the MAE was 0.488 years (95% confidence interval [CI]: 0.28–0.44; r2 = 0.978) for the AI algorithm and 0.771 years (95% CI: 0.64–0.90; r2 = 0.94) for the reader. In girls, the MAE was 0.494 years (95% CI: 0.41–0.56; r2 = 0.973) for the AI algorithm and 0.673 years (95% CI: 0.54–0.81; r2 = 0.934) for the reader. ConclusionThe AI solution better estimates the Greulich and Pyle bone age than a general radiologist does.

Characteristics of perivascular space dilatation in normal aging

The increased incidence of dilated perivascular spaces (dPVSs) visible on MRI has been observed with advancing age, but the relevance of PVS dilatation to normal aging across the lifespan has yet to be fully clarified. In the current study, we sought to find out the age dependence of dPVSs by exploring changes in different characteristics of PVS dilatation across a wide range of age. For 1220 healthy subjects aged between 18 and 100 years, PVSs were automatically segmented and characteristics of PVS dilatation were assessed in terms of the burden, location, and morphology of PVSs in the white matter (WM) and basal ganglia (BG). A machine learning model using the random forests method was constructed to estimate the subjects' age by employing the PVS features. The constructed machine learning model was able to estimate the age of the subjects with an error of 9.53 years on average (correlation=0.875). The importance of the PVS features indicated the primary contribution of the burden of PVSs in the BG and the additional contribution of locational and morphological changes of PVSs, specifically peripheral extension and reduced linearity, in the WM to age estimation. Indeed, adding the PVS location or morphology features to the PVS burden features provided an improvement to the performance of age estimation. The age dependence of dPVSs in terms of such various characteristics of PVS dilatation in healthy subjects could provide a more comprehensive reference for detecting brain disease-related PVS dilatation.

Radiographic dental age estimation applying and comparing Demirjian's seven (1973) and four (1976) teeth methods.

To apply Demirjian's seven (1973, D7) and four (1976, D4) teeth methods in a sample of Brazilian children and compare their performances for age estimation. Panoramic radiographs of 1200 children (600 females and 600 males) were collected from Southeast Brazil. The children were equally distributed (n = 60 females and 60 males) between age intervals of 1year between 6 and 15.9 (mean age = 10.9; standard deviation [SD]: 2.9) years. D7 and D4 dental age estimation methods were applied by two observers and were compared based on their error rates (error = estimated age-chronological age). Both methods overestimated the chronological age (p < 0.0001). D7 led to overestimations of 0.71 (p < 0.0001) in females and 0.83 (p < 0.0001) in males, while the overestimations with D4 were 0.14 (p = 0.0067) and 0.73 (p < 0.0001) for females and males, respectively. The mean age estimated by D7 was 11.7 (SD: 3.1) years, while for D4, it was 11.3 (SD: 3.3) years. Intra- and interobserver agreement was ≥ 0.98. D4 optimized the performance of radiographic dental age estimation compared to D7. Improvements were reflected in the total sample and separately for female and male children from Southeast Brazil.

Apparent age prediction from faces: A survey of modern approaches.

Apparent age estimation via human face image has attracted increased attention due to its numerous real-world applications. Predicting the apparent age has been quite difficult for machines and humans. However, researchers have focused on machine estimation of "age as perceived" to a high level of accuracy. To further improve the performance of apparent age estimation from the facial image, researchers continue to examine different methods to enhance its results further. This paper presents a critical review of the modern approaches and techniques for the apparent age estimation task. We also present a comparative analysis of the performance of some of those approaches on the apparent facial aging benchmark. The study also highlights the strengths and weaknesses of each approach used for apparent age estimation to guide in choosing the appropriate algorithms for future work in the field. The work focuses on the most popular algorithms and those that appear to have been the most successful for apparent age estimation to improve on the existing state-of-the-art results. We based our evaluations on three facial aging datasets, including looking at people (LAP)-2015, LAP-2016, and APPA-REAL, the most popular and publicly available datasets benchmark for apparent age estimation.

Age Estimation of Faces in Videos Using Head Pose Estimation and Convolutional Neural Networks.

Age estimation from human faces is an important yet challenging task in computer vision because of the large differences between physical age and apparent age. Due to the differences including races, genders, and other factors, the performance of a learning method for this task strongly depends on the training data. Although many inspiring works have focused on the age estimation of a single human face through deep learning, the existing methods still have lower performance when dealing with faces in videos because of the differences in head pose between frames, which can lead to greatly different results. In this paper, a combined system of age estimation and head pose estimation is proposed to improve the performance of age estimation from faces in videos. We use deep regression forests (DRFs) to estimate the age of facial images, while a multiloss convolutional neural network is also utilized to estimate the head pose. Accordingly, we estimate the age of faces only for head poses within a set degree threshold to enable value refinement. First, we divided the images in the Cross-Age Celebrity Dataset (CACD) and the Asian Face Age Dataset (AFAD) according to the estimated head pose degrees and generated separate age estimates for images with different poses. The experimental results showed that the accuracy of age estimation from frontal facial images was better than that for faces at different angles, thus demonstrating the effect of head pose on age estimation. Further experiments were conducted on several videos to estimate the age of the same person with his or her face at different angles, and the results show that our proposed combined system can provide more precise and reliable age estimates than a system without head pose estimation.

Applicability of Willems model of dental age estimation in a sample of southern Saudi Arabian children

ABSTRACT This study aimed to validate the Willems model in a southern Saudi Arabian sample to develop and validate an age estimation model specific for the southern Saudi Arabian population and compare both models’ age estimation performance. This cross-sectional study involved retrospective evaluation of orthopantomographs of 1206 children (658 males, 548 females) aged four to fourteen years. Initially, the Willems model was validated on the pooled sample. Next, this sample was split to develop a new Saudi-specific model using Willem’s approach. The age estimation performance of both models was compared. Differences between chronological age (CA) and estimated dental age (EDA) were calculated and expressed in mean error (ME), mean absolute error (MAE), and root mean square error (RMSE). The ME for the verified Willems model was 0.391 ± 1.479 years, whilst for the Saudi-specific model, it was 0.131 years ± 1.606 years. The difference in ME, MAE, and RMSE between the two models was 0.259, −0.113, and −0.082 years, respectively, and statistically significant (p < 0.05). However, clinically, these differences are trivial. Both the Willems model and the Saudi-specific model were similar in estimating CA. The use of the Willems model is recommended for estimating CA in the southern Saudi population.

Population-specific age estimation in Black Americans and Chinese people based on pulp chamber volume of first molars from cone beam computed tomography.

To validate the fitness of the age estimation model in Black Americans, which was previously and solely established for the Chinese population based on pulp chamber volume of the first molars from cone beam computed tomography (CBCT), and to establish a new age estimation model for Black Americans. A total of 203 subjects with CBCT scans, including 119 Chinese and 84 Black Americans, were retrospectively identified. The age range of subjects was between 11 and 87years. For both populations, automated 3D pulp chamber segmentation of the first molars was performed by deep learning, followed by volume calculation and age estimation by a logarithmic regression model, which was established in a prior study solely on Chinese population. Additionally, a separate logarithmic regression analysis was carried out on Black Americans. The performance of age estimation was assessed by the mean absolute error (MAE), root mean square error (RMSE), Wilcoxon signed rank test, and coefficient of determination (R2) between the actual and estimated human ages. When applying the age estimation model established in the prior study, MAE = 7.994years and RMSE = 10.065years were observed in the Chinese population, while MAE = 14.049years and RMSE = 17.866years were observed in Black Americans. The new age estimation model established for Black Americans was AGE = 89.752 - 21.176 × lnV (V = pulp chamber volume), with MAE = 7.930years, RMSE = 10.664years, and coefficient of determination (R2) = 0.600. Population-specific age estimation is needed when applied in Black Americans and Chinese people based on pulp chamber volume of the first molars from CBCT.

LAE-GAN-Based Face Image Restoration for Low-Light Age Estimation

Age estimation is applicable in various fields, and among them, research on age estimation using human facial images, which are the easiest to acquire, is being actively conducted. Since the emergence of deep learning, studies on age estimation using various types of convolutional neural networks (CNN) have been conducted, and they have resulted in good performances, as clear images with high illumination were typically used in these studies. However, human facial images are typically captured in low-light environments. Age information can be lost in facial images captured in low-illumination environments, where noise and blur generated by the camera in the captured image reduce the age estimation performance. No study has yet been conducted on age estimation using facial images captured under low light. In order to overcome this problem, this study proposes a new generative adversarial network for low-light age estimation (LAE-GAN), which compensates for the brightness of human facial images captured in low-light environments, and a CNN-based age estimation method in which compensated images are input. When the experiment was conducted using the MORPH, AFAD, and FG-NET databases—which are open databases—the proposed method exhibited more accurate age estimation performance and brightness compensation in low-light images compared to state-of-the-art methods.

Deep Learning Based Real Age and Gender Estimation from Unconstrained Face Image towards Smart Store Customer Relationship Management

The COVID-19 pandemic markedly changed the human shopping nature, necessitating a contactless shopping system to curb the spread of the contagious disease efficiently. Consequently, a customer opts for a store where it is possible to avoid physical contacts and shorten the shopping process with extended services such as personalized product recommendations. Automatic age and gender estimation of a customer in a smart store strongly benefit the consumer by providing personalized advertisement and product recommendation; similarly, it aids the smart store proprietor to promote sales and develop an inventory perpetually for the future retail. In our paper, we propose a deep learning-founded enterprise solution for smart store customer relationship management (CRM), which allows us to predict the age and gender from a customer’s face image taken in an unconstrained environment to facilitate the smart store’s extended services, as it is expected for a modern venture. For the age estimation problem, we mitigate the data sparsity problem of the large public IMDB-WIKI dataset by image enhancement from another dataset and perform data augmentation as required. We handle our classification tasks utilizing an empirically leading pre-trained convolutional neural network (CNN), the VGG-16 network, and incorporate batch normalization. Especially, the age estimation task is posed as a deep classification problem followed by a multinomial logistic regression first-moment refinement. We validate our system for two standard benchmarks, one for each task, and demonstrate state-of-the-art performance for both real age and gender estimation.

Utilizing transfer learning of pre-trained AlexNet and relevance vector machine for regression for predicting healthy older adult’s brain age from structural MRI

Discrepancies between the estimated brain age from brain structural MRI and the chronological age have been associated with a broad spectrum of neurocognitive disorders. The performance of brain age estimation heavily depends on predefined or hand-crafted features. Although 3D convolutional neural network (CNN) based approaches have been proposed, they require high computational cost, large memory load, and numerous images. Coupling a pre-trained 2D CNN for transfer learning with a well-established relevance vector machine for regression approach can greatly enhance the capabilities of the model. Several important strategies, including feature transfer learning, 3D feature concatenation, and dimensionality reduction were taken. The estimated brain age was modeled by structural magnetic resonance imaging (sMRI) from 594 normal healthy older individuals (age 50–90 years). We proposed and manifested a pre-trained AlexNet as a robust feature extractor. Also, the considerable cost of developing a 3D CNN was avoided by applying 3D feature concatenation and data reduction. The proposed method achieves superior performance with a mean absolute error of 4.51 years for old subjects. The predicted brain age also demonstrated high test-retest reliability (intra-class correlation coefficient of 0.979). The effectiveness and robustness of the proposed model were well studied. The proposed approach can compete with or even outperform those state-of-the-art approaches, and feature transfer learning strategy can introduce new perspectives to some well-established brain age prediction models with predefined or hand-crafted features.

Vec2UAge: Enhancing underage age estimation performance through facial embeddings

Automated facial age estimation has drawn increasing attention in recent years. Several applications relevant to digital forensic investigations include the identification of victims, suspects and missing children, and the decrease of investigators’ exposure to psychologically impacting material. Nevertheless, due to the lack of accurately labelled age datasets, particularly for the underage age range, sufficient performance accuracy remains a major challenge in the field of age estimation. To address the problem, a novel regression-based model was created, Vec2UAge. FaceNet embeddings were extracted and used as feature vectors to train the model from the VisAGe and Selfie-FV datasets. A balanced, unbiased dataset was created for testing and validation. Data augmentation techniques were evaluated to further be used to expand the training dataset. The learning rate (lr) is one of the most important hyper-parameters for deep neural networks; a cyclic learning rate approach was used to find the optimal initial value for lr and the performance was evaluated. The distribution of model performance was presented per optimiser and one of the winning models with a Stochastic Weight Averaging (SWA) optimised training run reached a mean absolute error rate as low as 2.36 years. Additionally, the time of convergence using SWA was significantly faster than other optimisers evaluated, i.e., ADAGRAD, ADAM and Stochastic Gradient Descent. The evaluation model metric is presented in a form of a distribution rather than a single value, giving more insights into the effects of the random initialisations, optimisers and the learning rate on the outcome.

Application of Cameriere’s method for dental age estimation in children in South China

The aim of this study was to evaluate the applicability of Cameriere’s European formula for age estimation in children in South China and to adapt the formula to establish a more suitable formula for these children. Moreover, the performance of dental age estimation based on Cameriere’s method combining the developmental information of permanent teeth (PT) and third molar (TM) was also analysed. Orthopantomographs of 720 healthy children in Group A, and orthopantomographs of 320 children and 280 subadults in Group B were assessed. The samples of Group A were divided into training dataset 1 and test dataset 1, and the samples of Group B were also divided into training dataset 2 and test dataset 2. A South China-specific formula was established based on the training dataset 1, and the comparison of accuracy between the Cameriere’s European formula and the South China-specific formula was conducted with the test dataset 1. Additionally, a PT regression model, a TM regression model, and a combined regression model (PT + TM) were established based on the training dataset 2, and the performance of these three models were validated on the test dataset 2. The Cameriere’s European formula underestimated chronological age with a mean difference (ME) of −0.47 ± 1.11 years in males and −0.69 ± 1.19 years in females. However, the South China-specific formula underestimated chronological age, with a mean difference (ME) of −0.02 ± 0.71 years in males and −0.14 ± 0.73 years in females. Compared with PT model and TM model, the PT and TM combined model obtained the smallest root mean square error (RMSE) of 1.29 years in males and 0.93 years in females. In conclusion, the South China-specific formula was more suitable for assessing the dental age of children in South China, and the PT and TM combined model can improve the accuracy of dental age estimation in children. Key points Orthopantomographs of 720 healthy children in Group A, and orthopantomographs of 320 children and 280 subadults in Group B were assessed. A South China-specific formula was established based on the training dataset 1, and the comparison of accuracy between the Cameriere’s European formula and the South China-specific formula was conducted with the test dataset 1. A PT regression model, a TM regression model, and a combined regression model (PT + TM) were established based on the training dataset 2, and the performance of these three models were validated on the test dataset 2. The South China-specific formula was more suitable for assessing the dental age of children in South China, and the PT and TM combined model can improve the accuracy of dental age estimation in children.

Estimation of age from speech using excitation source features

Extraction of various distinct features from speech, Gaussian mixture models (GMMs) have been employed as classifiers. Age estimation performance has been inferred by employing excitation source features (LPCCs). The detected age of a speaker is deemed to belong to any of the 9 age groups of 5–10, 10–15, 15–20, 20–25, 25–30, 30–35, 35–40, 40–45 and 45–50. An age group speech corpus has been collected in this work by recording voice of speakers of Hindi language of different age groups and dialects using ten text prompts in neutral speech. Textually neutral Hindi words have been used to construct text prompts which have been recorded in neutral emotion. Different age group has been characterized by using these texts prompt. The average age performance of multispeaker (male + female) is around 94%. In this research, classification of different group of speakers has been done on the basis of excitation source in human speech.

Distribution Cognisant Loss for Cross-Database Facial Age Estimation With Sensitivity Analysis.

Existing facial age estimation studies have mostly focused on intra-database protocols that assume training and test images are captured under similar conditions. This is rarely valid in practical applications, where we typically encounter training and test sets with different characteristics. In this article, we deal with such situations, namely subjective-exclusive cross-database age estimation. We formulate the age estimation problem as the distribution learning framework, where the age labels are encoded as a probability distribution. To improve the cross-database age estimation performance, we propose a new loss function which provides a more robust measure of the difference between ground-truth and predicted distributions. The desirable properties of the proposed loss function are theoretically analysed and compared with the state-of-the-art approaches. In addition, we compile a new balanced large-scale age estimation database. Last, we introduce a novel evaluation protocol, called subject-exclusive cross-database age estimation protocol, which provides meaningful information of a method in terms of the generalisation capability. The experimental results demonstrate that the proposed approach outperforms the state-of-the-art age estimation methods under both intra-database and subject-exclusive cross-database evaluation protocols. In addition, in this article, we provide a comparative sensitivity analysis of various algorithms to identify trends and issues inherent to their performance. This analysis introduces some open problems to the community which might be considered when designing a robust age estimation system.

Magnetic resonance imaging for forensic age estimation in living children and young adults: a systematic review.

The use of MRI in forensic age estimation has been explored extensively during the last decade. The authors of this paper synthesized the available MRI data for forensic age estimation in living children and young adults to provide a comprehensive overview that can guide age estimation practice and future research. To do so, the authors searched MEDLINE, Embase and Web of Science, along with cited and citing articles and study registers. Two authors independently selected articles, conducted data extraction, and assessed risk of bias. They considered study populations including living subjects up to 30years old. Fifty-five studies were included in qualitative analysis and 33 in quantitative analysis. Most studies had biases including use of relatively small European (Caucasian) populations, varying MR approaches and varying staging techniques. Therefore, it was not appropriate to pool the age distribution data. The authors found that reproducibility of staging was remarkably lower in clavicles than in any other anatomical structure. Age estimation performance was in line with the gold standard, radiography, with mean absolute errors ranging from 0.85years to 2.0years. The proportion of correctly classified minors ranged from 65% to 91%. Multifactorial age estimation performed better than that based on a single anatomical site. The authors found that more multifactorial age estimation studies are necessary, together with studies testing whether the MRI data can safely be pooled. The current review results can guide future studies, help medical professionals to decide on the preferred approach for specific cases, and help judicial professionals to interpret the evidential value of age estimation results.

Hierarchical age estimation mechanism with adaBoost-based deep instance weighted fusion

ABSTRACT Age estimation can obtain biological age which is helpful for diagnosis of healthy status and disease. The current age estimation methods do not consider the deep relationships of instances, which limits the potential improvement of the age estimation performance. A hierarchical age estimation mechanism with adaboost-based deep instance weighted fusion is proposed to solve this problem. First, a circulation iterative means clustering (CIMC) algorithm is designed for constructing the hierarchical instance space (multiple-layer instance spaces) and obtain multiple trained base regression models. Second, an adaboost-based deep instance weighted fusion (ADIWF) mechanism is designed to fuse the results of the trained regression models. Several representative age-related datasets are used for verification of the proposed method. The experimental results show that the mean absolute error (MAE) can be decreased apparently, by 6.86% and 1.42% on the Heart and Diabetes Dataset, respectively. Besides, some factors that may influence the performance of the proposed mechanism are studied. In general, the proposed age estimation mechanism is effective. In addition, the mechanism is a kind of framework mechanism, so it can be used to construct different concrete age estimation algorithms, and is helpful for related studies.

DeepUAge: Improving Underage Age Estimation Accuracy to Aid CSEM Investigation

Age is a soft biometric trait that can aid law enforcement in the identification of victims of Child Sexual Exploitation Material (CSEM) creation/distribution. Accurate age estimation of subjects can classify explicit content possession as illegal during an investigation. Automation of this age classification has the potential to expedite content discovery and focus the investigation of digital evidence through the prioritisation of evidence containing CSEM. In recent years, artificial intelligence based approaches for automated age estimation have been created, and many public cloud service providers offer this service on their platforms. The accuracy of these algorithms have been improving over recent years. These existing approaches perform satisfactorily for adult subjects, but perform wholly inadequately for underage subjects.To this end, the largest underage facial age dataset, VisAGe, has been used in this work to train a ResNet50 based deep learning model, DeepUAge, that achieved state-of-the-art beating performance for age estimation of minors. This paper describes the design and implementation of this model. An evaluation, validation and comparison of the proposed model is performed against existing facial age classifiers resulting in the best overall performance for underage subjects.