Article
Computer Science, Artificial Intelligence
Ram Krishn Mishra, Siddhaling Urolagin, J. Angel Arul Jothi, Pramod Gaur
Summary: Image processing is a technique used to apply various operations to images to improve them or extract information, with facial recognition being a prominent application. This study examines the accuracy of categorizing human facial expressions using deep learning and transfer learning methods, proposing a deep hybrid learning approach that combines multiple deep learning models.
IMAGE AND VISION COMPUTING
(2022)
Article
Computer Science, Information Systems
Na Liu, Fan Zhang, Fuqing Duan
Summary: In this paper, a method is proposed that performs personalized local feature extraction and builds hierarchical age features by iteratively erasing local regions of interest. A joint multi-input and multi-output network is designed to learn age classification and regression tasks, using global features and personalized local features as inputs. Extensive experiments validate the effectiveness of the proposed method for age estimation.
JOURNAL OF VISUAL COMMUNICATION AND IMAGE REPRESENTATION
(2023)
Article
Computer Science, Artificial Intelligence
Bin-Bin Gao
Summary: This paper investigates the use of label distribution learning in ordinal regression tasks such as facial age and attractiveness estimation, particularly through deep label distribution learning (DLDL) methods integrated into deep convolutional neural networks. Existing DLDL methods suffer from inconsistency between training objectives and evaluation metrics, resulting in suboptimal performance. Additionally, these methods tend to employ image classification or face recognition models with a high number of parameters, leading to expensive computation cost and storage overhead. This paper firstly analyzes the relationship between two state-of-the-art methods - ranking CNN and DLDL - and demonstrates that the ranking method inherently learns label distribution. It unifies these two popular methods within the DLDL framework. Furthermore, a lightweight network architecture and a unified framework are proposed to address the inconsistency issue and reduce resource consumption. The effectiveness of this approach is demonstrated on facial age and attractiveness estimation tasks, achieving state-of-the-art results with 36 times fewer parameters and 3 times faster inference speed.
NEURAL COMPUTING & APPLICATIONS
(2023)
Article
Psychiatry
Jie Huang, Yanli Zhao, Wei Qu, Zhanxiao Tian, Yunlong Tan, Zhiren Wang, Shuping Tan
Summary: This study proposes a rapid detection method for schizophrenia based on deep learning and facial videos. The results show that the method can differentiate between healthy controls and schizophrenic patients by analyzing changes in facial area, providing assistance for diagnosis in a clinical setting.
ASIAN JOURNAL OF PSYCHIATRY
(2022)
Article
Environmental Sciences
Sichen Wang, Yanfeng Huo, Xi Mu, Peng Jiang, Shangpei Xun, Binfang He, Wenyu Wu, Lin Liu, Yonghong Wang
Summary: A novel deep learning model based on convolutional neural network architecture was developed to estimate ground-level ozone concentrations in eastern China. The model demonstrated high accuracy and robustness compared to in situ measurements and other machine learning techniques. This study provides an efficient and exact method for estimating ground-level ozone and offers a new perspective for modeling spatiotemporal air pollutants.
Article
Computer Science, Information Systems
Nazir Shabbir, Ranjeet Kumar Rout
Summary: This paper proposes a unique facial expression recognition system that improves performance by incorporating different variant patterns in facial images. The proposed model utilizes a deep convolutional neural network and incorporates cultural difference datasets. Experimental results show that the FER system achieves more significant performance compared to other state-of-the-art models.
MULTIMEDIA TOOLS AND APPLICATIONS
(2023)
Article
Computer Science, Information Systems
Huiying Zhang, Yu Zhang, Xin Geng
Summary: This study proposes a more practical approach for facial age recognition, which limits the age label distribution to only cover a reasonable number of neighboring ages, and explores different label distributions to improve model performance. The experimental results show that the proposed method is more effective for facial age recognition compared to the current state-of-the-art framework DLDL.
FRONTIERS OF COMPUTER SCIENCE
(2021)
Article
Ecology
Yu Qi, Han Su, Rong Hou, Hangxing Zang, Peng Liu, Mengnan He, Ping Xu, Zhihe Zhang, Peng Chen
Summary: The conservation of giant pandas has attracted great attention, and determining their age groups accurately is important. Traditional methods have limitations, but this study developed a deep learning method based on facial images to achieve high accuracy in age group classification of captive giant pandas.
ECOLOGY AND EVOLUTION
(2022)
Article
Computer Science, Artificial Intelligence
Denis Milosevic, Marin Vodanovic, Ivan Galic, Marko Subasic
Summary: This study explores the applicability of deep learning for chronological age estimation, determines the best performing model parameters using pretrained general-purpose vision model parameters as the starting point, and highlights the importance of different anatomical regions of the dental system for estimation through ablation experiments. The proposed approach achieves the lowest estimation error in literature for adult and senior subjects, verified on one of the largest datasets of panoramic dental x-ray images, setting a baseline for future research in forensic odontology.
EXPERT SYSTEMS WITH APPLICATIONS
(2022)
Article
Environmental Studies
Stuart J. Barnes, Samuel Nathan Kirshner
Summary: The study found that trust and attractiveness based on host's facial features contribute to nearly a 5% increase in prices for Airbnb accommodation. Additionally, trust is more important in situations of smaller accommodation shared with strangers.
TOURISM MANAGEMENT
(2021)
Article
Agriculture, Dairy & Animal Science
Junbin Liu, Deqin Xiao, Youfu Liu, Yigui Huang
Summary: Constructing a contactless pig mass estimation method using computer vision technology can improve pig breeding program and production efficiency. The developed deep learning-based pig mass estimation model can quickly and accurately estimate the body mass of pigs in an unconstrained environment, providing real-time evaluation of body quality for grading and adjusting breeding plans.
Article
Computer Science, Information Systems
M. A. H. Akhand, Shuvendu Roy, Nazmul Siddique, Md Abdus Samad Kamal, Tetsuya Shimamura
Summary: This study proposes a highly accurate facial emotion recognition system based on a very deep CNN model, utilizing transfer learning technique for system construction and optimization. The method shows remarkable accuracy and superiority on two different facial image datasets, demonstrating proficiency in handling diverse profile views and frontal views.
Article
Computer Science, Artificial Intelligence
Yuanyuan Shang, Yuchen Pan, Xiao Jiang, Zhuhong Shao, Guodong Guo, Tie Liu, Hui Ding
Summary: In this paper, a method called LQGDNet is proposed to combine the advantages of hand-crafted and deep features for depression recognition. This method is the first attempt to use a quaternion-based method for facial depression recognition and shows superior performance compared to existing methods in experiments.
IEEE TRANSACTIONS ON AFFECTIVE COMPUTING
(2023)
Article
Computer Science, Information Systems
Ying Wang, Ping Liu, Jinrui Feng, Lijun Wang
Summary: Human body pose estimation is a technology used to locate the key points of the human body in various industries and sports. This paper proposes a method that utilizes grid convolutional coding neural network to address the limitations faced when using regression heat maps for key point localization. By dividing the image into grids and accurately locating the key points, high precision positioning is achieved, and a multi-person posture estimation algorithm is proposed for scenarios involving multiple individuals.
HUMAN-CENTRIC COMPUTING AND INFORMATION SCIENCES
(2023)
Article
Computer Science, Information Systems
Hamid Sadeghi, Abolghasem-A Raie
Summary: This paper proposes a deep histogram metric learning approach based on a Convolutional Neural Network (CNN) for facial expression recognition. By introducing a histogram calculation layer and a learnable matrix, the accuracy is improved and the proposed CNN outperforms state-of-the-art methods on multiple databases.
INFORMATION SCIENCES
(2022)
Article
Radiology, Nuclear Medicine & Medical Imaging
Olubunmi Sule, Serestina Viriri
Summary: This paper proposes an improved method for optimal segmentation of blood vessels in retinal fundus images using convolutional neural networks (CNNs). By enhancing the contrast of the RGB and green channel, the improved images are evaluated for quality using various measures. The results show that the improved RGB quality outperforms the improved green channel, indicating that using RGB for contrast enhancement effectively improves the image quality. The proposed method achieves an accuracy of 94.47%, sensitivity of 70.92%, specificity of 98.20%, and AUC (ROC) of 97.56% on the DRIVE dataset.
JOURNAL OF DIGITAL IMAGING
(2023)
Review
Computer Science, Theory & Methods
Adekanmi Adeyinka Adegun, Serestina Viriri, Jules-Raymond Tapamo
Summary: This research evaluates and analyzes the performance of deep learning approaches, including Convolutional Neural Networks and vision transformer, for classification of high-resolution satellite images. Various CNN-based models were explored and evaluated on publicly available datasets. The results demonstrate the feasibility of Deep Learning approaches in learning the complex features of remote sensing images.
JOURNAL OF BIG DATA
(2023)
Review
Imaging Science & Photographic Technology
Stewart Muchuchuti, Serestina Viriri
Summary: Millions of people worldwide suffer from retinal abnormalities, and early detection and treatment are crucial for preventing avoidable blindness. Manual disease detection is time-consuming, tedious, and lacks consistency. Efforts have been made to automate ocular disease detection using Deep Convolutional Neural Networks (DCNNs) and vision transformers (ViTs) for Computer-Aided Diagnosis (CAD). However, the complex nature of retinal lesions presents challenges. This work reviews common retinal pathologies, imaging modalities, and deep-learning research for the detection and grading of various retinal diseases, concluding that CAD through deep learning will play an increasingly vital role in assisting healthcare professionals.
JOURNAL OF IMAGING
(2023)
Article
Remote Sensing
Zubair Saeed, Muhammad Haroon Yousaf, Rehan Ahmed, Sergio A. Velastin, Serestina Viriri
Summary: Object detection is challenging with aerial images due to small target sizes, low resolution, occlusion, attitude, and scale variations. We modified the CenterNet architecture and used different CNN-based backbones to improve performance. The modified CenterNet achieved promising results on challenging datasets and was compared with other popular object detectors. The approach was also optimized and implemented on edge platforms, showing good performance compared to the latest cutting-edge research on both discrete GPU and edge platforms.
Article
Chemistry, Analytical
Adekanmi Adeyinka Adegun, Jean Vincent Fonou Dombeu, Serestina Viriri, John Odindi
Summary: Object detection in high-resolution remote sensing satellite images is critical for various purposes, including disaster prevention, service delivery, and urban/rural planning. This study evaluated the performance of deep learning-based object detection methods on a new dataset of diverse features. The results showed that YOLOv8 achieved the highest detection accuracy of more than 90% and the fastest detection speed of 0.2 ms.
Article
Multidisciplinary Sciences
Samson Akinpelu, Serestina Viriri
Summary: Speech emotion classification has become very important in recent years and plays a significant role in Human-Computer Interaction and affective computing. This study proposes an attention-based network combining pre-trained convolutional neural network and regularized neighbourhood component analysis for improved classification of speech emotion. The proposed model achieves better performance compared to other state-of-the-art approaches.
SCIENTIFIC REPORTS
(2023)
Article
Chemistry, Multidisciplinary
Abubaker Abdelrahman, Serestina Viriri
Summary: This paper proposes a new deep learning model (SE-ResNet) for segmentation of kidney tumors, which accurately identifies and segments regions of kidneys and tumors in CT images, achieving high performance. This is of great importance for early diagnosis and timely intervention, improving patient outcomes.
APPLIED SCIENCES-BASEL
(2023)
Article
Imaging Science & Photographic Technology
Wilson Bakasa, Serestina Viriri
Summary: This study utilizes cutting-edge deep learning techniques to identify pancreatic ductal adenocarcinoma (PDAC) using computerized tomography (CT) medical imaging. The proposed hybrid model, VGG16-XGBoost, performs well on PDAC images, achieving an accuracy and weighted F1 score of 0.97 for the dataset under study. The results of this study are extremely helpful for PDAC diagnosis from CT pancreas images, categorizing them into different TNM staging system class labels (T0, T1, T2, T3, and T4).
JOURNAL OF IMAGING
(2023)
Article
Computer Science, Information Systems
Samson Akinpelu, Serestina Viriri, Adekanmi Adegun
Summary: This article introduces an efficient lightweight model for speech emotion recognition, which integrates Random Forest and Multilayer Perceptron classifiers into the VGGNet framework. The experimental results show that the proposed model achieves high recognition accuracy of 100%, 96%, and 86.25% on the TESS, EMODB, and RAVDESS datasets, respectively, surpassing the recent state-of-the-art model found in the literature.
Article
Computer Science, Information Systems
Tirivangani Magadza, Serestina Viriri
Summary: Brain tumors are a major cause of death in adults, and accurate and timely segmentation of brain tumors is critical for treatment planning and disease monitoring. This study proposes a method that achieves high-quality tumor segmentation with lower computational cost by utilizing a modified network architecture and introducing an attention mechanism.
Article
Computer Science, Theory & Methods
Adam Hassan, Serestina Viriri
Summary: This article introduces an improved method for facial recognition by resizing each facial component to extract invariant features and improve cross-age facial recognition. The experimental results show better accuracy compared to related research on facial databases.
INTERNATIONAL JOURNAL OF ADVANCED COMPUTER SCIENCE AND APPLICATIONS
(2022)
Proceedings Paper
Computer Science, Software Engineering
Samson Akinpelu, Serestina Viriri
Summary: The significant role of emotion in human daily interaction cannot be over-emphasized. This study proposes a deep transfer learning model for speech emotion classification, which shows improved results in accuracy and specificity.
ADVANCES IN VISUAL COMPUTING, ISVC 2022, PT II
(2022)
Article
Computer Science, Information Systems
Tirivangani Magadza, Serestina Viriri
Summary: This study proposes an efficient network architecture for brain tumor segmentation, which partially utilizes depthwise separable convolutions to reduce computational costs. The experimental results show comparable performance with the state-of-the-art methods while minimizing computational complexity. A critical analysis of current efficient model designs is also provided.
Article
Imaging Science & Photographic Technology
Abubaker Abdelrahman, Serestina Viriri
Summary: Deep learning models play a crucial role in kidney tumor segmentation, assisting clinicians in accurately identifying and segmenting tumors, and improving the efficacy of tumor treatment.
JOURNAL OF IMAGING
(2022)
Article
Computer Science, Information Systems
Jane Oruh, Serestina Viriri, Adekanmi Adegun
Summary: In this study, an enhanced deep learning LSTM recurrent neural network (RNN) model was proposed to address the limitation of traditional LSTM in processing continuous input streams. The proposed model incorporates RNN as a forget gate in the memory block to reset cell states, enabling more efficient processing of continuous input streams.