Search

Gait recognition is the identification of individuals based on how they walk. It can identify an individual of interest without their intervention, making it better suited for surveillance from afar. Computer-aided silhouette-based gait analysis is frequently employed due to its efficiency and effectiveness. However, covariate conditions have a significant influence on individual recognition because they conceal essential features that are helpful in recognizing individuals from their walking style. To address such issues, we proposed a novel deep-learning framework to tackle covariate conditions in gait by proposing regions subject to covariate conditions. The features extracted from those regions will be neglected to keep the model's performance effective with custom kernels. The proposed technique sets aside static and dynamic areas of interest, where static areas contain covariates, and then features are learnt from the dynamic regions unaffected by covariates to effectively recognize individuals. The features were extracted using three customized kernels, and the results were concatenated to produce a fused feature map. Afterward, CNN learns and extracts the features from the proposed regions to recognize an individual. The suggested approach is an end-to-end system that eliminates the requirement for manual region proposal and feature extraction, which would improve gait-based identification of individuals in real-world scenarios. The experimentation is performed on publicly available dataset i.e. CASIA A, and CASIA C. The findings indicate that subjects wearing bags produced 90 % accuracy, and subjects wearing coats produced 58 % accuracy. Likewise, recognizing individuals with different walking speeds also exhibited excellent results, with an accuracy of 94 % for fast and 96 % for slow-paced walk patterns, which shows improvement compared to previous deep learning methods.© 2017 Elsevier Inc. All rights reserved.

Deep learning models perform well when there is enough data available for training, but otherwise the performance deteriorates rapidly owing to the so-called data shortage problem. Recently, model-agnostic meta-learning (MAML) was proposed to alleviate this problem by embedding common prior knowledge from different tasks into the initial parameters of the target model. Data shortages are very common in regional influenza predictions, and MAML also often struggles with regional influenza forecasting, especially when region-specific knowledge, such as peak timing or intensity, varies. In this paper, we propose a novel MAML-based parameter adjustment scheme for influenza forecasting, called MARAPAS. The fundamental idea of our scheme is to adjust the initial parameters obtained from common knowledge to a target region by using adjustment variables. We experimentally show that MARAPAS outperforms other MAML-based methods, in terms of root mean square error and Pearson correlation coefficient. Particularly, this scheme improves the forecasting performance by up to 34 % compared with that of the state-of-the-art schemes. We also show the robust forecasting accuracy of our scheme and demonstrate its applicability by performing zero-shot COVID-19 forecasting.

Although a generative adversarial network (GAN) can generate realistic and distinct images, it requires numerous training data. Data augmentation is a popular method of incrementing data using various augmentation operations. However, data augmentation for training GANs can lead to augmentation leaks where unintended distortion effects in augmented images appear in images generated by the GAN. This augmentation leaks occur because the GAN is trained to generate images similar to the training dataset containing augmented images. Although recent studies have revealed that maintaining the invertibility of data augmentation can prevent augmentation leaks, maintaining it for various data augmentation configurations and datasets is challenging. This paper proposes a novel auxiliary neural network called an auxiliary augmentation operation classifier (AAOC) to prevent augmentation leaks. The AAOC learns the classification of augmentation operations during GAN training, guiding the GAN to avoid generating images with unintended distortions. Due to its simple and lightweight architecture, the AAOC does not increase the training time of GANs. Extensive experiments on the three public datasets revealed that the AAOC reduced the occurrence of augmentation leaks by a factor of 30 compared to the basic method and improved image quality by 50% for leak-free images.

Deep learning models that require vast amounts of training data struggle to achieve good animal sound classification (ASC) performance. Among recent few-shot ASC methods to address the data shortage problem regarding animals that are difficult to observe, model-agnostic meta-learning (MAML) has shown new possibilities by encoding common prior knowledge derived from different tasks into the model parameter initialization of target tasks. However, when the knowledge on animal sounds is difficult to generalize due to its diversity, MAML exhibits poor ASC performance due to the static initialization setting. In this paper, we propose a novel task-adaptive parameter transformation scheme called few-shot ASC. TAPT generates transformation variables while learning common knowledge and uses the variables to make parameters specific to the target task. Owing to this transformation, TAPT can reduce overfitting and enhance adaptability, training speed, and performance in heterogeneous tasks compared to MAML. In experiments on two public datasets on the same backbone network, we show that TAPT outperforms the existing few-shot ASC schemes in terms of classification accuracy, and in particular a performance improvement of 20.32% compared to the state-of-the-art scheme. In addition, we show that TAPT is robust to hyperparameters and efficient for training.

The human face is one of the most viewed visual objects in a person's life and is used for identifying a person through facial landmarks, which includes the eyes, nose, mouth, and ears that make up a face. It is also possible to communicate nonverbally through the movements of facial landmarks; that is, change of facial expression. Thus, facial landmarks play a crucial role in human-related image analysis. Automatic facial landmark detection is a challenging problem in the field of computer vision, and various studies are underway. The emergence of Deep Neural Networks has played an important role in solving difficult problems in computer vision. Semantic segmentation is a field in which images are classified into pixel units and has also developed rapidly by incorporating deep learning. In this paper, we propose a method for accurately extracting facial landmarks using semantic segmentation. First, we introduce a semantic segmentation architecture for sophisticated landmark detection, and datasets composed of facial images and ground truth pairs. Then, we suggest how improve the performance of pixel classification by adjusting the imbalance of the number of pixels according to the face landmark. Through extensive experiments, we evaluated our approach using the metrics pixel accuracy and intersection over union.

Load forecasting is one of the critical tasks for enhancing the energy efficiency of smart grids. Even though recent deep learning-based load forecasting models have shown excellent forecasting performance, one of the common problems they faced was that their forecasting accuracy was highly dependent on the data quality and quantity available for the model training. Collecting a sufficient amount of high-quality data is expensive and time-consuming. Recently, a generative adversarial network (GAN) has shown its potential as a solution to the data shortage problem by generating virtual data based on a small amount of real data, and several studies have used GAN to generate electric load data for training forecasting models. However, due to the noise data problem of GANs, their predictive performance also deteriorated. To solve this problem, in this study, we propose a two-stage data generation scheme that more effectively generates input and output variables for short-term load forecasting. In the first stage, we generate virtual calendar and temperature data used as input variables using a conditional tabular GAN (CTGAN). In the second stage, we generate electric load data corresponding to the input variables using a deep learning-based regression model. Lastly, we construct our forecasting model by training another regression model using a mixture of generated data and real data. To verify the effectiveness of our scheme, we conducted extensive experiments using various datasets and data generation models. We report some of the results.

Recently, energy procurement by renewable energy sources has increased. In particular, as solar power generation has a high penetration rate among them, solar radiation predictions at the site are attracting much attention for efficient operation. Various approaches have been proposed to forecast solar radiation accurately. Recently, hybrid models have been proposed to improve performance through forecasting in the frequency domain using past solar radiation. Since solar radiation data have a pattern, forecasting in the frequency domain can be effective. However, forecasting performance deteriorates on days when the weather suddenly changes. In this paper, we propose a domain hybrid forecasting model that can respond to weather changes and exhibit improved performance. The proposed model consists of two stages. In the first stage, forecasting is performed in the frequency domain using wavelet transform, complete ensemble empirical mode decomposition, and multilayer perceptron, while forecasting in the sequence domain is accomplished using light gradient boosting machine. In the second stage, a multilayer perceptron-based domain hybrid model is constructed using the forecast values of the first stage as the input. Compared with the frequency-domain model, our proposed model exhibits an improvement of up to 36.38% in the normalized root-mean-square error.

Animal sound classification (ASC) refers to the automatic identification of animal categories by sound, and is useful for monitoring rare or elusive wildlife. Thus far, deep-learning-based models have shown good performance in ASC when training data is sufficient, but suffer from severe performance degradation if not. Recently, generative adversarial networks (GANs) have shown the potential to solve this problem by generating virtual data. However, in a multi-class environment, existing GAN-based methods need to construct separate generative models for each class. Additionally, they only consider the waveform or spectrogram of sound, resulting in poor quality of the generated sound. To overcome these shortcomings, we propose a two-step sound augmentation scheme using a class-conditional GAN. First, common features are learned from all classes of animal sounds, and multiple classes of animal sounds are generated based on the features that consider both waveforms and spectrograms using class-conditional GAN. Second, we select data from the generated data based on the confidence of the pretrained ASC model to improve classification performance. Through experiments, we show that the proposed method improves the accuracy of the basic ASC model by up to 18.3%, which corresponds to a performance improvement of 13.4% compared to the second-best augmentation method.

Abstract As a variety of video surveillance devices such as CCTV, drones, and car dashboard cameras have become popular, numerous studies have been conducted regarding the effective enforcement of security and surveillance based on video analysis. In particular, in car-related surveillance, car tracking is the most challenging task. One early approach to accomplish such a task was to analyze frames from different video sources separately. Considering the shooting range of the bulk of video devices, the outcome from the analysis of single video source is highly limited. To obtain more comprehensive information for car tacking, a set of video sources should be considered together and the relevant information should be integrated according to spatial and temporal constraints. Therefore, in this study, we propose a real-time car tracking system based on surveillance videos from diverse devices including CCTV, dashboard cameras, and drones. For scalability and fault tolerance, our system is built on a distributed processing framework and comprises a Frame Distributor, a Feature Extractor, and an Information Manager. The Frame Distributor is responsible for distributing the video frames from various devices to the processing nodes. The Feature Extractor extracts principal vehicle features such as plate number, location, and time from each frame. The Information Manager stores all the features into a database and handles user requests by collecting relevant information from the feature database. To illustrate the effectiveness of our proposed system, we implemented a prototype system and performed a number of experiments. We report some of the results.

An efficient energy operation strategy for the smart grid requires accurate day-ahead electricity load forecasts with high time resolutions, such as 15 or 30 min. Most high-time resolution electricity load prediction techniques deal with a single output prediction, so their ability to cope with sudden load changes is limited. Multistep-ahead forecasting addresses this problem, but conventional multistep-ahead prediction models suffer from deterioration in prediction performance as the prediction range is expanded. In this paper, we propose a novel two-stage multistep-ahead forecasting model that combines a single-output forecasting model and a multistep-ahead forecasting model to solve the aforementioned problem. In the first stage, we perform a single-output prediction based on recent electricity load data using a light gradient boosting machine with time-series cross-validation, and feed it to the second stage. In the second stage, we construct a multistep-ahead forecasting model that applies an attention mechanism to sequence-to-sequence bidirectional long short-term memory (S2S ATT-BiLSTM). Compared to the single S2S ATT-BiLSTM model, our proposed model achieved improvements of 3.23% and 4.92% in mean absolute percentage error and normalized root mean square error, respectively.

Smart grids have recently attracted increasing attention because of their reliability, flexibility, sustainability, and efficiency. A typical smart grid consists of diverse components such as smart meters, energy management systems, energy storage systems, and renewable energy resources. In particular, to make an effective energy management strategy for the energy management system, accurate load forecasting is necessary. Recently, artificial neural network–based load forecasting models with good performance have been proposed. For accurate load forecasting, it is critical to determine effective hyperparameters of neural networks, which is a complex and time-consuming task. Among these parameters, the type of activation function and the number of hidden layers are critical in the performance of neural networks. In this study, we construct diverse artificial neural network–based building electric energy consumption forecasting models using different combinations of the two hyperparameters and compare their performance. Experimental results indicate that neural networks with scaled exponential linear units and five hidden layers exhibit better performance, on average than other forecasting models.

Species distribution models have been used for various purposes, such as conserving species, discovering potential habitats, and obtaining evolutionary insights by predicting species occurrence. Many statistical and machine-learning-based approaches have been proposed to construct effective species distribution models, but with limited success due to spatial biases in presences and imbalanced presence-absences. We propose a novel species distribution model to address these problems based on bootstrap aggregating (bagging) ensembles of deep neural networks (DNNs). We first generate bootstraps considering presence-absence data on spatial balance to alleviate the bias problem. Then we construct DNNs using environmental data from presence and absence locations, and finally combine these into an ensemble model using three voting methods to improve prediction accuracy. Extensive experiments verified the proposed model’s effectiveness for species in South Korea using crowdsourced observations that have spatial biases. The proposed model achieved more accurate and robust prediction results than the current best practice models.

Recently, multistep-ahead prediction has attracted much attention in electric load forecasting because it can deal with sudden changes in power consumption caused by various events such as fire and heat wave for a day from the present time. On the other hand, recurrent neural networks (RNNs), including long short-term memory and gated recurrent unit (GRU) networks, can reflect the previous point well to predict the current point. Due to this property, they have been widely used for multistep-ahead prediction. The GRU model is simple and easy to implement; however, its prediction performance is limited because it considers all input variables equally. In this paper, we propose a short-term load forecasting model using an attention based GRU to focus more on the crucial variables and demonstrate that this can achieve significant performance improvements, especially when the input sequence of RNN is long. Through extensive experiments, we show that the proposed model outperforms other recent multistep-ahead prediction models in the building-level power consumption forecasting.

Smart islands have focused on renewable energy sources, such as solar and wind, to achieve energy self-sufficiency. Because solar photovoltaic (PV) power has the advantage of less noise and easier installation than wind power, it is more flexible in selecting a location for installation. A PV power system can be operated more efficiently by predicting the amount of global solar radiation for solar power generation. Thus far, most studies have addressed day-ahead probabilistic forecasting to predict global solar radiation. However, day-ahead probabilistic forecasting has limitations in responding quickly to sudden changes in the external environment. Although multistep-ahead (MSA) forecasting can be used for this purpose, traditional machine learning models are unsuitable because of the substantial training time. In this paper, we propose an accurate MSA global solar radiation forecasting model based on the light gradient boosting machine (LightGBM), which can handle the training-time problem and provide higher prediction performance compared to other boosting methods. To demonstrate the validity of the proposed model, we conducted a global solar radiation prediction for two regions on Jeju Island, the largest island in South Korea. The experiment results demonstrated that the proposed model can achieve better predictive performance than the tree-based ensemble and deep learning methods.

Biodiversity conservation is important for the protection of ecosystems. One key task for sustainable biodiversity conservation is to effectively preserve species’ habitats. However, for various reasons, many of these habitats have been reduced or destroyed in recent decades. To deal with this problem, it is necessary to effectively identify potential habitats based on habitat suitability analysis and preserve them. Various techniques for habitat suitability estimation have been proposed to date, but they have had limited success due to limitations in the data and models used. In this paper, we propose a novel scheme for assessing habitat suitability based on a two-stage ensemble approach. In the first stage, we construct a deep neural network (DNN) model to predict habitat suitability based on observations and environmental data. In the second stage, we develop an ensemble model using various habitat suitability estimation methods based on observations, environmental data, and the results of the DNN from the first stage. For reliable estimation of habitat suitability, we utilize various crowdsourced databases. Using observational and environmental data for four amphibian species and seven bird species in South Korea, we demonstrate that our scheme provides a more accurate estimation of habitat suitability compared to previous other approaches. For instance, our scheme achieves a true skill statistic (TSS) score of 0.886, which is higher than other approaches (TSS = 0.725 ± 0.010).

Electric energy consumption forecasting is an interesting, challenging, and important issue in energy management and equipment efficiency improvement. Existing approaches are predictive models that have the ability to predict for a specific profile, i.e., a time series of a whole building or an individual household in a smart building. In practice, there are many profiles in each smart building, which leads to time-consuming and expensive system resources. Therefore, this study develops a robust framework for the Multiple Electric Energy Consumption forecasting (MEC) of a smart building using Transfer Learning and Long Short-Term Memory (TLL), the so-called MEC-TLL framework. In this framework, we first employ a k-means clustering algorithm to cluster the daily load demand of many profiles in the training set. In this phase, we also perform Silhouette analysis to specify the optimal number of clusters for the experimental datasets. Next, this study develops the MEC training algorithm, which utilizes a cluster-based strategy for transfer learning the Long Short-Term Memory models to reduce the computational time. Finally, extensive experiments are conducted to compare the computational time and different performance metrics for multiple electric energy consumption forecasting on two smart buildings in South Korea. The experimental results indicate that our proposed approach is capable of economical overheads while achieving superior performances. Therefore, the proposed approach can be applied effectively for intelligent energy management in smart buildings.

Facial landmarks represent prominent feature points on the face that can be used as anchor points in many face-related tasks. So far, a lot of research has been done with the aim of achieving efficient extraction of landmarks from facial images. Employing a large number of feature points for landmark detection and tracking usually requires excessive processing time. On the contrary, relying on too few feature points cannot accurately represent diverse landmark properties, such as shape. To extract the 68 most popular facial landmark points efficiently, in our previous study, we proposed a model called EMTCNN that extended the multi-task cascaded convolutional neural network for real-time face landmark detection. To improve the detection accuracy, in this study, we augment the EMTCNN model by using two convolution techniques—dilated convolution and CoordConv. The former makes it possible to increase the filter size without a significant increase in computation time. The latter enables the spatial coordinate information of landmarks to be reflected in the model. We demonstrate that our model can improve the detection accuracy while maintaining the processing speed.

For efficient and effective energy management, accurate energy consumption forecasting is required in energy management systems (EMSs). Recently, several artificial intelligence-based techniques have been proposed for accurate electric load forecasting; moreover, perfect energy consumption data are critical for the prediction. However, owing to diverse reasons, such as device malfunctions and signal transmission errors, missing data are frequently observed in the actual data. Previously, many imputation methods have been proposed to compensate for missing values; however, these methods have achieved limited success in imputing electric energy consumption data because the period of data missing is long and the dependency on historical data is high. In this study, we propose a novel missing-value imputation scheme for electricity consumption data. The proposed scheme uses a bagging ensemble of multilayer perceptrons (MLPs), called softmax ensemble network, wherein the ensemble weight of each MLP is determined by a softmax function. This ensemble network learns electric energy consumption data with explanatory variables and imputes missing values in this data. To evaluate the performance of our scheme, we performed diverse experiments on real electric energy consumption data and confirmed that the proposed scheme can deliver superior performance compared to other imputation methods.

We introduce a distance metric between two distributions and propose a Generative Adversarial Network (GAN) model: the Simplified Fréchet distance (SFD) and the Simplified Fréchet GAN (SFGAN). Although the data generated through GANs are similar to real data, GAN often undergoes unstable training due to its adversarial structure. A possible solution to this problem is considering Fréchet distance (FD). However, FD is unfeasible to realize due to its covariance term. SFD overcomes the complexity so that it enables us to realize in networks. The structure of SFGAN is based on the Boundary Equilibrium GAN (BEGAN) while using SFD in loss functions. Experiments are conducted with several datasets, including CelebA and CIFAR-10. The losses and generated samples of SFGAN and BEGAN are compared with several distance metrics. The evidence of mode collapse and/or mode drop does not occur until 3000k steps for SFGAN, while it occurs between 457k and 968k steps for BEGAN. Experimental results show that SFD makes GANs more stable than other distance metrics used in GANs, and SFD compensates for the weakness of models based on BEGAN-based network structure. Based on the experimental results, we can conclude that SFD is more suitable for GAN than other metrics.

An energy-management system requires accurate prediction of the electric load for optimal energy management. However, if the amount of electric load data is insufficient, it is challenging to perform an accurate prediction. To address this issue, we propose a novel electric load forecasting scheme using the electric load data of diverse buildings. We first divide the electric energy consumption data into training and test sets. Then, we construct multivariate random forest (MRF)-based forecasting models according to each building except the target building in the training set and a random forest (RF)-based forecasting model using the limited electric load data of the target building in the test set. In the test set, we compare the electric load of the target building with that of other buildings to select the MRF model that is the most similar to the target building. Then, we predict the electric load of the target building using its input variables via the selected MRF model. We combine the MRF and RF models by considering the different electric load patterns on weekdays and holidays. Experimental results demonstrate that combining the two models can achieve satisfactory prediction performance even if the electric data of only one day are available for the target building.

Smart grid systems, which have gained much attention due to its ability to reduce operation and management costs of power systems, consist of diverse components including energy storage, renewable energy, and combined cooling, heating and power (CCHP) systems. The CCHP has been investigated to reduce energy costs by using the thermal energy generated during the power generation process. For efficient utilization of CCHP and numerous power generation systems, accurate short-term load forecasting (STLF) is necessary. So far, even though many single algorithm-based STLF models have been proposed, they showed limited success in terms of applicability and coverage. This problem can be alleviated by combining such single algorithm-based models in ways that take advantage of their strengths. In this paper, we propose a novel two-stage STLF scheme; extreme gradient boosting and random forest models are executed in the first stage, and deep neural networks are executed in the second stage to combine them. To show the effectiveness of our proposed scheme, we compare our model with other popular single algorithm-based forecasting models and then show how much electric charges can be saved by operating CCHP based on the schedules made by the economic analysis on the predicted electric loads.

The electric energy consumption prediction (EECP) is an essential and complex task in intelligent power management system. EECP plays a significant role in drawing up a national energy development policy. Therefore, this study proposes an Electric Energy Consumption Prediction model utilizing the combination of Convolutional Neural Network (CNN) and Bi-directional Long Short-Term Memory (Bi-LSTM) that is named EECP-CBL model to predict electric energy consumption. In this framework, two CNNs in the first module extract the important information from several variables in the individual household electric power consumption (IHEPC) dataset. Then, Bi-LSTM module with two Bi-LSTM layers uses the above information as well as the trends of time series in two directions including the forward and backward states to make predictions. The obtained values in the Bi-LSTM module will be passed to the last module that consists of two fully connected layers for finally predicting the electric energy consumption in the future. The experiments were conducted to compare the prediction performances of the proposed model and the state-of-the-art models for the IHEPC dataset with several variants. The experimental results indicate that EECP-CBL framework outperforms the state-of-the-art approaches in terms of several performance metrics for electric energy consumption prediction on several variations of IHEPC dataset in real-time, short-term, medium-term and long-term timespans.

Observing animal movements enables us to understand animal behavior changes, such as migration, interaction, foraging, and nesting. Based on spatiotemporal changes in weather and season, animals instinctively change their position for foraging, nesting, or breeding. It is known that moving patterns are closely related to their traits. Analyzing and predicting animals’ movement patterns according to spatiotemporal change offers an opportunity to understand their unique traits and acquire ecological insights into animals. Hence, in this paper, we propose an animal movement prediction scheme using a predictive recurrent neural network architecture. To do that, we first collect and investigate geo records of animals and conduct pattern refinement by using random forest interpolation. Then, we generate animal movement patterns using the kernel density estimation and build a predictive recurrent neural network model to consider the spatiotemporal changes. In the experiment, we perform various predictions using 14 K long-billed curlew locations that contain their five-year movements of the breeding, non-breeding, pre-breeding, and post-breeding seasons. The experimental results confirm that our predictive model based on recurrent neural networks can be effectively used to predict animal movement.

Besides genetic characteristics, people also undergo a process of skin aging under the influence of diverse factors such as sun exposure, food intake, sleeping patterns, and drinking habits, which are closely related to their personal lifestyle. So far, many studies have been conducted to analyze skin conditions quantitatively. However, to describe the current skin condition or predict future skin aging effectively, we need to understand the correlation between skin aging and lifestyle. In this study, we first demonstrate how to trace people’s skin condition accurately using scale-invariant feature transform and the color histogram intersection method. Then, we show how to estimate skin texture aging depending on the lifestyle by considering various features from face, neck, and hand dermoscopy images. Lastly, we describe how to predict future skin conditions in terms of skin texture features. Based on the Pearson correlation, we describe the correlation between skin aging and lifestyle, and estimate skin aging according to lifestyle using the polynomial regression and support vector regression models. We evaluate the performance of our proposed scheme through various experiments.

A stable power supply is very important in the management of power infrastructure. One of the critical tasks in accomplishing this is to predict power consumption accurately, which usually requires considering diverse factors, including environmental, social, and spatial-temporal factors. Depending on the prediction scope, building type can also be an important factor since the same types of buildings show similar power consumption patterns. A university campus usually consists of several building types, including a laboratory, administrative office, lecture room, and dormitory. Depending on the temporal and external conditions, they tend to show a wide variation in the electrical load pattern. This paper proposes a hybrid short-term load forecast model for an educational building complex by using random forest and multilayer perceptron. To construct this model, we collect electrical load data of six years from a university campus and split them into training, validation, and test sets. For the training set, we classify the data using a decision tree with input parameters including date, day of the week, holiday, and academic year. In addition, we consider various configurations for random forest and multilayer perceptron and evaluate their prediction performance using the validation set to determine the optimal configuration. Then, we construct a hybrid short-term load forecast model by combining the two models and predict the daily electrical load for the test set. Through various experiments, we show that our hybrid forecast model performs better than other popular single forecast models.

Facial wrinkles are widely used to evaluate skin condition or aging for various fields such as skin diagnosis, plastic surgery consultations, and cosmetic recommendations. In order to effectively process facial wrinkles in facial image analysis, accurate wrinkle segmentation is required to identify wrinkled regions. Existing deep learning-based methods have difficulty segmenting fine wrinkles due to insufficient wrinkle data and the imbalance between wrinkle and non-wrinkle data. Therefore, in this paper, we propose a new facial wrinkle segmentation method based on a UNet++ model. Specifically, we construct a new facial wrinkle dataset by manually annotating fine wrinkles across the entire face. We then extract only the skin region from the facial image using a facial landmark point extractor. Lastly, we train the UNet++ model using both dice loss and focal loss to alleviate the class imbalance problem. To validate the effectiveness of the proposed method, we conduct comprehensive experiments using our facial wrinkle dataset. The experimental results showed that the proposed method was superior to the latest wrinkle segmentation method by 9.77%p and 10.04%p in IoU and F1 score, respectively. [ABSTRACT FROM AUTHOR]