Prof Bhekisipho Twala
@tut.ac.za
Deputy Vice-Chancellor (Digital Transformation)
Tshwane University of Technology
Bhekisipho Twala is the deputy vice-chancellor for digital transformation and professor of artificial intelligence and data science at the Tshwane University of Technology in South Africa. Before then he was the Executive Dean of Engineering and Built Environment at the Durban University of Technology (DUT), South Africa. Before joining DUT, he was the Director of the School of Engineering at the University of South Africa and the founder of the Institute for Intelligent Systems at the University of Johannesburg, South Africa. He completed his PhD at the Open University (UK) in 2005. He was a post-doctoral researcher at Brunel University in the UK, mainly focussing on applying machine learning in software engineering. Prof Twala’s current research includes promoting and conducting research in Artificial Intelligence within the Big Data Analytics field and developing novel and innovative solutions to crucial research problems.
EDUCATION
BA (Economics & Statistics)
Post Graduate Certificate (Statistics)
MSc (Statistics)
PhD (Artificial Intelligence & Statistical Science)
RESEARCH, TEACHING, or OTHER INTERESTS
Multidisciplinary, Statistics, Probability and Uncertainty, Statistics and Probability, Electrical and Electronic Engineering
FUTURE PROJECTS
Red Cards and Red Flags: Understanding Domestic Violence in Football Communities
"Red Cards and Red Flags: Understanding Domestic Violence in Football Communities" explores the intricate relationship between football culture and the prevalence of domestic violence. This study investigates how the intense emotions and cultural norms associated with football can contribute to domestic violence incidents, particularly in communities where football is deeply ingrained. Through a combination of qualitative and quantitative research methods, including interviews, surveys, and analysis of domestic violence reports during football seasons, the study seeks to identify patterns and underlying causes. Key findings suggest that heightened emotions during football matches, alcohol consumption, and deeply rooted gender norms within football communities can exacerbate domestic violence. The study also highlights how the competitive and aggressive nature of the sport can spill over into personal relationships, leading to increased tension and conflict at home.
Applications Invited
Masters, PhD students & Collaboration
The Psychological Impact of "Indoda Ayikhayi"
The cultural norm encapsulated in the phrase "Indoda ayikhali" (a man does not cry) exerts a profound psychological impact on men's mental health by promoting emotional repression. This study delves into how societal expectations for men to suppress their emotions contribute to adverse mental health outcomes. Using a combination of psychological research, case studies, and empirical data, we explore the consequences of this cultural expectation on men's mental well-being. The findings reveal that the suppression of emotions, as dictated by the "Indoda ayikhayi" ethos, often leads to increased levels of stress, anxiety, and depression among men. This cultural norm discourages men from seeking emotional support and expressing vulnerability, resulting in a reluctance to engage in mental health services. The study also highlights the role of societal and familial reinforcement in perpetuating this behaviour, creating a cycle of emotional repression that spans generations.
Applications Invited
Masters, PhD students & Collaboration
Utilizing Machine Learning Algorithms to Enhance Climate Models.
Integrating Artificial Intelligence (AI) in optimizing renewable energy systems significantly advances the fight against climate change. This study explores how AI can enhance the efficiency and effectiveness of solar and wind energy systems, which are pivotal to transitioning towards sustainable energy sources. AI algorithms, such as machine learning and neural networks, are employed to predict energy production based on weather patterns, optimize renewable energy infrastructure maintenance schedules, and manage energy storage and distribution in smart grids. By analyzing vast amounts of meteorological data, AI can accurately forecast solar and wind energy outputs, thereby improving the reliability and stability of renewable energy supplies. AI-driven predictive maintenance reduces downtime and prolongs the lifespan of energy systems, ensuring continuous and efficient operation.
Applications Invited
Masters, PhD students & Collaboration
Scopus Publications
Scholar Citations
Scholar h-index
Scholar i10-index
Scopus Publications
Sanjay Chauhan, Rajesh Singh, Anita Gehlot, Shaik Vaseem Akram, Bhekisipho Twala, and Neeraj Priyadarshi
MDPI AG
In paper [...]
Bhekisipho Twala
Routledge
Shailendra Tiwari, Anita Gehlot, Rajesh Singh, Bhekisipho Twala, and Neeraj Priyadarshi
Elsevier BV
Mlungisi Duma, Bhekisipho Twala, and Tshilidzi Marwala
Elsevier BV
Bhekisipho Twala
Frontiers Media SA
BackgroundParkinson’s disease (PD) represents one of the most prevalent neurodegenerative disorders globally, affecting over 10 million individuals worldwide. Traditional diagnostic approaches rely heavily on clinical observation and subjective assessment, often leading to delayed or inaccurate diagnoses. The emergence of artificial intelligence (AI) technologies offers unprecedented opportunities for precision diagnosis and personalized treatment strategies in PD management.ObjectiveThis study aims to comprehensively review current AI applications in Parkinson’s disease diagnosis and treatment, evaluate existing methodologies, and present experimental results from a novel multimodal AI diagnostic framework.MethodsA systematic review was conducted across PubMed, IEEE Xplore, and Web of Science databases from 2018 to 2024, focusing on AI applications in PD diagnosis and treatment. Additionally, we developed and tested a hybrid machine learning model combining deep learning, computer vision, and natural language processing techniques for PD assessment using motor symptom analysis, voice pattern recognition, and gait analysis.ResultsThe systematic review identified 127 relevant studies demonstrating significant advances in AI-driven PD diagnosis, with accuracy rates ranging from 78 to 96%. Our experimental framework achieved 94.2% accuracy in early-stage PD detection, outperforming traditional clinical assessment methods. The integrated approach showed particular strength in identifying subtle motor fluctuations and predicting treatment response patterns.ConclusionAI-driven approaches demonstrate substantial potential for revolutionizing PD diagnosis and treatment personalization. The integration of multiple data modalities and advanced machine learning algorithms enables earlier detection, more accurate monitoring, and optimized therapeutic interventions. Future research should focus on large-scale clinical validation and implementation frameworks for healthcare systems.
Pawan Kumar Pathak, Anil Kumar Yadav, Sanjeevikumar Padmanaban, Bhekisipho Twala, and Innocent Kamwa
Institution of Engineering and Technology (IET)
AbstractThis paper proposes an intelligent battery charging scheme for hybrid electric vehicles (HEVs) with a fuel cell as the primary energy source and solar photovoltaic (PV) and battery as the auxiliary energy sources. While dealing with the PV, a minimized oscillation‐based improved perturb and observe (I‐P&O) maximum power point (MPP) tracking (MPPT) scheme is designed to mitigate the impact of oscillations around MPP and loss of tracking direction. The DC–DC boost and DC–DC buck power converters are connected in a cascade manner to harvest optimal power from PV and as a charging circuit for HEV, respectively. An intelligent fuzzy logic‐based proportional integral derivative (PID) (F‐PID) controller is employed for the buck converter to get the constant voltage and constant current for the effective charging of the battery. The two primary objectives of this work are (1) maximum utilization of the designed PV array via the I‐P&O MPPT scheme to enhance the system efficacy, reduce system cost, and reduce complexity. (2) To obtain minimum battery losses and an enhanced life cycle of HEV. The proposed MPPT scheme provides a maximum 99.80% tracking efficiency of the considered PV array at an insolation level of 1000 W/m2. Moreover, almost nominal voltage and current ripples have appeared in HEV's proposed intelligent battery charging circuit.
Sanjay Singh Chauhan, Pradeep Suri, Bhekisipho Twala, Neeraj Priyadarshi, and Farman Ali
F1000 Research Ltd
Background of the study The influence of macroeconomic indicators makes it important to study the relationship between macroeconomic indicators and stock market return. On further analysis it can be observed that different sectors respond differently to change in the macroeconomic indicator that is important for investors, researchers and policy makers. Methods The autoregressive distributed lag (ARDL) model is applied to study influence of macroeconomic indicators on sectoral return of NSE from April 2012 to August 2024. Results Findings of the study show that macroeconomic indicators influence sectoral return in the short run as well as long run and the influence is differential. The analysis of long run relationship shows that Foreign Institutional Investment (FII) significantly affects all the sectoral indices except IT. Index of industrial production (IIP) have significant relationship with Auto, IT, Media, Metal and Pharma. Money supply (MS) significantly affects Bank, FMCG and IT in the long run. Wholesale Price Index (WPI) has significant relationship with Auto, FMCG and Media in the long run. Economic Policy Uncertainty Index (EPU) affects Auto, FMCG and Pharma in the long run. Crude oil price (COP) has significant effect only on Media in the long run. Exchange rate (ER) does not have significant effect on any of the sectoral index. Conclusion In the long run FII, IIP, EPU, MS and WIP are major determinants of stock market return. In the short run FII, ER and COP are major determinants of stock market return.
Sumet Mehta, Fei Han, Muhammad Sohail, Bhekisipho Twala, Asad Ullah, Fasee Ullah, Arfat Ahmad Khan, and Qinghua Ling
PeerJ
The analysis of high-dimensional microarray gene expression data presents critical challenges, including excessive dimensionality, increased computational burden, and sensitivity to random initialization. Traditional optimization algorithms often produce inconsistent and suboptimal results, while failing to preserve local data structures limiting both predictive accuracy and biological interpretability. To address these limitations, this study proposes an adaptive neighborhood-preserving multi-objective particle swarm optimization (ANPMOPSO) framework for gene selection. ANPMOPSO introduces four key innovations: (1) a weighted neighborhood-preserving ensemble embedding (WNPEE) technique for dimensionality reduction that retains local structure; (2) Sobol sequence (SS) initialization to enhance population diversity and convergence stability; (3) a differential evolution (DE)-based adaptive velocity update to dynamically balance exploration and exploitation; and (4) a novel ranking strategy that combines Pareto dominance with neighborhood preservation quality to prioritize biologically meaningful gene subsets. Experimental evaluations on six benchmark microarray datasets and eleven multi-modal test functions (MMFs) demonstrate that ANPMOPSO consistently outperforms state-of-the-art methods. For example, it achieves 100% classification accuracy on Leukemia and Small-Round-Blue-Cell Tumor (SRBCT) using only 3–5 genes, improving accuracy by 5–15% over competitors while reducing gene subsets by 40–60%. Additionally, on MMFs, ANPMOPSO attains superior hypervolume values (e.g., 1.0617 ± 0.2225 on MMF1, approximately 10–20% higher than competitors), confirming its robustness in balancing convergence and diversity. Although the method incurs higher training time due to its structural and adaptive components, it achieves a strong trade-off between computational cost and biological relevance, making it a promising tool for high-dimensional gene selection in bioinformatics.
Ramesh Chandra Pathak, Priyam Agarwal, Anita Gehlot, Rajesh Singh, Shaik Vaseem Akram, Lovi Raj Gupta, Amit Kumar Thakur, Neeraj Priyadarshi, and Bhekisipho Twala
Bentham Science Publishers Ltd.
Culture and Tourism are two mainly interrelated elements that contribute a lot to achieving Sustainable Development for any developing country especially India, which has an extremely rich historical and cultural background. Tourism Industry is the fastest growing sector in a local economy creating several job opportunities which ultimately raise the standard of living of people which further raises the consumption level of goods and services, resulting in a rise in the Gross Domestic Product (GDP) of a country. However, various studies pointed out major promotional strategies concerning tourism and culture but an amalgamated promotional approach for both was still missing. With this motivation, the current study aims at providing an amalgamated promotional approach in assimilation with the latest Industry 4.0 technologies such as Artificial Intelligence (AI), Machine Learning (ML), Big Data, Blockchain, Virtual Reality (VR), Digital Twin and Metaverse to the Indian tourism industry by reviewing prior research studies. The findings of the current study are establishing an online future travel demands forecasting system, an online tourists’ destination personalized recommendation system, an online tourist’s review analysis recommendation system, and an online destination image recommendation system and provide the practical design for it through 1+5 Architectural Views Model and by applying several ML algorithms such as CNN, BPNN, SVM, Collaborative Filtering, K-means Clustering, API Emotion, and Naïve Bayes algorithms. Finally, this study has discussed challenges and suggested vital recommendations for future work with the assimilation of Industry 4.0 technologies.
Abhishek Anand, Muhamad Mansor, Kamal Sharma, Amritanshu Shukla, Atul Sharma, Md Irfanul Haque Siddiqui, Kishor Kumar Sadasivuni, Neeraj Priyadarshi, and Bhekisipho Twala
Elsevier BV
Gurmit Singh, Simant Kamal Dutta, Rajesh Singh, Anita Gehlot, Dharam Buddhi, Siddharth Swami, Shaik Vaseem Akram, Amit Kumar Thakur, Neeraj Priyadarshi, and Bhekisipho Twala
Elsevier BV
Shailendra Tiwari, Anita Gehlot, Rajesh Singh, Bhekisipho Twala, and Neeraj Priyadarshi
Elsevier BV
Ashish Singh Chauhan, Rajesh Singh, Neeraj Priyadarshi, Bhekisipho Twala, Surindra Suthar, and Siddharth Swami
Springer Science and Business Media LLC
AbstractThis study explores the practical applications of artificial intelligence (AI) in medical imaging, focusing on machine learning classifiers and deep learning models. The aim is to improve detection processes and diagnose diseases effectively. The study emphasizes the importance of teamwork in harnessing AI’s full potential for image analysis. Collaboration between doctors and AI experts is crucial for developing AI tools that bridge the gap between concepts and practical applications. The study demonstrates the effectiveness of machine learning classifiers, such as forest algorithms and deep learning models, in image analysis. These techniques enhance accuracy and expedite image analysis, aiding in the development of accurate medications. The study evidenced that technologically assisted medical image analysis significantly improves efficiency and accuracy across various imaging modalities, including X-ray, ultrasound, CT scans, MRI, etc. The outcomes were supported by the reduced diagnosis time. The exploration also helps us to understand the ethical considerations related to the privacy and security of data, bias, and fairness in algorithms, as well as the role of medical consultation in ensuring responsible AI use in healthcare.
Vikrant Pachouri, Awadhesh Chandramauli, Rajesh Singh, Anita Gehlot, Neeraj Priyadarshi, and Bhekisipho Twala
Springer Science and Business Media LLC
AbstractAccording to the United Nations, Sustainable Development Goals (SDG) goal 6 and goal 14 seeks to ensure the sustainable management of water present over the earth for all. Urban cities saw a lot of expansion in terms of population and the number of industries established there. Water quality analysis becomes a huge requirement of today’s scenario due to the impurities present in water which harms the livelihood of society. Some of the hazardous impurities like heavy metals destruct the surroundings. In this study, the removal of heavy metals from wastewater with an efficient technique i.e. Bioremediation is represented with the analysis and evaluation of results recorded in the laboratory testing. Two samples were taken from two different sites which are being analyzed with the implementation methodology represented in the present article. The inclusion of the ANOVA model for the optimization of the outcome generated is evaluated and received the P-value and F-crit value. Two algae Chlorella Minutissima & Chlorella Singularis were evaluated based on their impurity removal efficiency as well an analysis of the biological treatment method over other chemical methods has been examined. The results were analyzed and represented in the form of a table as well and the variation in the value of WQP is shown in the form of graphs. The observation shows that the variation in WQP after the integration of algae lies under the permissible limit. Alkalinity is estimated in the range of 20–40 mg/l, Hardness lies in the range of 0–60mg/l, and pH comes approximately in a range of 6.5–8. The results of the ANOVA model is also depicted in graphical form highlighting the P-value and F-crit value of different result generated. Finally, the summary of the proposed work is illustrated with the challenges faced and future recommendations have been provided. Based on the evaluation, the framework is generated for the efficient technique used for heavy metal removal i.e. Bioremediation which provides a great advancement in the efficacy of removal of heavy metals.
Vishant Kumar, Rajesh Singh, Anita Gehlot, Shaik Vaseem Akram, Amit Kumar Thakur, Ronald Aseer, Neeraj Priyadarshi, and Bhekisipho Twala
Springer Science and Business Media LLC
AbstractBlack ice is responsible for dangerous road-related incidents that can cause collisions and harm vehicle drivers and pedestrians. Visual examination and weather forecasts are two standard traditional methods for detecting black ice on roads, but they are often inaccurate and may not deliver the vehicle driver with up-to-date information on road conditions. The evolution of Industry 4.0 enabling technologies such as wireless sensor network (WSN), Internet of Things (IoT), cloud computing, and machine learning (ML) has been capable of detecting events in real time. This study aims to analyse the integration of the WSN, IoT, ML, and image processing for black ice detection. The qualitative research method is followed in this study, where the problems of black ice detection are studied. Following this, the role of Industry 4.0 enabling technologies is analyzed in detail for black ice detection. According to the study, we can detect black ice using different methods, but some methods need to be refined if we talk about the prediction. By merging different technologies, we can improve the overall architecture and create an algorithm that works with images and physical variables like temperature, humidity, due point, and road temperature, which were responsible for black ice formation, and predict the chances of black ice formation by training the system.
G. Gopichand, T. Sarath, Ankur Dumka, Himanshu Rai Goyal, Rajesh Singh, Anita Gehlot, Lovi Raj Gupta, Amit Kumar Thakur, Neeraj Priyadarshi, and Bhekisipho Twala
Springer Science and Business Media LLC
Hitesh Bhatt, Rajesh Bahuguna, Siddharth Swami, Rajesh Singh, Anita Gehlot, Shaik Vaseem Akram, Lovi Raj Gupta, Amit Kumar Thakur, Neeraj Priyadarshi, and Bhekisipho Twala
Springer Science and Business Media LLC
AbstractThe judiciary is a foundation of democracy, upholding the rule of law and protecting rights. Efficient court administration is vital for public trust, justice, and timely proceedings. Currently, courts face challenges such as inconsistencies, adjournments, and absence of transparency, undermining the justice system. Traditional manual processes and paper-based documentation cause significant backlogs, slow resolutions, and limited public access. As case volumes and complexities rise, modernizing court administration through digital transformation is progressively critical. Currently, many countries are integrating technologies in the courts for its administration and other activities. In recent years, courts and judges have been subjected to pressure to improve performance, uplifting judicial effectiveness to a top priority. Subsequently, several countries have integrated simplification and digitization strategies in judicial initiatives to enhance court efficiency. Switzerland’s Justitia 4.0 and Brazil’s PJE are notable initiatives that focused to strengthen court administration through digitalization. These aspects motivated this study to examine the detailed integration of industry 4.0 technologies such as the Internet of things, cloud computing, blockchain, big data analytics, robotics, drones, Metaverse, digital twins, and artificial intelligence for court administration with digitalized infrastructure. According to the study, integrating these technologies in less complex cases helps minimize expenditures and save time, making to resolve cases conveniently, efficiently, and superiorly. The study also identified the challenges and issues associated with industry 4.0 technologies such as evidence gathering, evidence preservation, robot judges for pre-judgment analysis, and judgment delivery, which future studies need to be explored.
Dumisani Selby Nkambule, Bhekisipho Twala, and Jan Harm Christiaan Pretorius
MDPI AG
Credit risk is a crucial component of daily financial services operations; it measures the likelihood that a borrower will default on a loan, incurring an economic loss. By analysing historical data for assessment of the creditworthiness of a borrower, lenders can reduce credit risk. Data are vital at the core of the credit decision-making processes. Decision-making depends heavily on accurate, complete data, and failure to harness high-quality data would impact credit lenders when assessing the loan applicants’ risk profiles. In this paper, an empirical comparison of the robustness of seven machine learning algorithms to credit risk, namely support vector machines (SVMs), naïve base, decision trees (DT), random forest (RF), gradient boosting (GB), K-nearest neighbour (K-NN), and logistic regression (LR), is carried out using the Lending Club credit data from Kaggle. This task uses seven performance measures, including the F1 Score (recall, accuracy, and precision), ROC-AUC, and HL and MCC metrics. Then, the harnessing of generative adversarial networks (GANs) simulation to enhance the robustness of the single machine learning classifiers for predicting credit risk is proposed. The results show that when GANs imputation is incorporated, the decision tree is the best-performing classifier with an accuracy rate of 93.01%, followed by random forest (92.92%), gradient boosting (92.33%), support vector machine (90.83%), logistic regression (90.76%), and naïve Bayes (89.29%), respectively. The classifier is the worst-performing method with a k-NN (88.68%) accuracy rate. Subsequently, when GANs are optimised, the accuracy rate of the naïve Bayes classifier improves significantly to (90%) accuracy rate. Additionally, the average error rate for these classifiers is over 9%, which implies that the estimates are not far from the actual values. In summary, most individual classifiers are more robust to missing data when GANs are used as an imputation technique. The differences in performance of all seven machine learning algorithms are significant at the 95% level.
Prafful Negi, Ashish Pathani, Bhuvan Chandra Bhatt, Siddharth Swami, Rajesh Singh, Anita Gehlot, Amit Kumar Thakur, Lovi Raj Gupta, Neeraj Priyadarshi, Bhekisipho Twala,et al.
MDPI AG
The incorporation of Industry 4.0 has integrated various innovations into fire safety management, thus changing the mode of identifying, assessing, and controlling fire risks. This review aims at how emerging technologies like IoT, AI, cloud technology, and BIM are making changes to fire safety in structural structures. With IoT-enabled sensors, data, and analytics coupled with predictive algorithms for real-time scenarios, fire safety systems have become dynamic systems where early detection, quick response, and risk management can be achieved. In addition, cloud web-based solutions improve the storage of information while providing the predictive aspect for certainty of safety measures. This paper also largely focuses on such activities through the likes of ISO/IEC 30141 and IEEE 802.15.4, thus making a critical role in maintaining effective connectivity between IoT devices, which is necessary for the effective performance of fire safety systems. Furthermore, the implementation issues, including the high costs, the difficulty in scaling up the projects, and the cybersecurity concerns, are considered and compared to the possible solutions, which include upgrading in stages and the possibility of subsidies from the government. The review also points out areas for further study, such as the creation of small cell networks with lower latency, the use of AI to carry out the maintenance of IoTs, and the enhancement of protection mechanisms of systems that are based on the IoTs. In general, this paper highlights the vast possibilities offered by Industry 4.0 technologies to support organizational fire safety management or decrease fire fatalities and improve built environment fire safety.
Gopal Krishna, Rajesh Singh, Anita Gehlot, Vaseem Akram Shaik, Bhekisipho Twala, and Neeraj Priyadarshi
Elsevier BV
Prafful Negi, Gaurav Thakur, Rajesh Singh, Anita Gehlot, Amit Kumar Thakur, Lovi Raj Gupta, Neeraj Priyadarshi, and Bhekisipho Twala
Elsevier BV
Sibusiso Reuben Bakana, Yongfei Zhang, and Bhekisipho Twala
Elsevier BV
Yashwant Singh Bisht, Rajesh Singh, Anita Gehlot, Shaik Vaseem Akram, Amit Kumar Thakur, Neeraj Priyadarshi, and Bhekisipho Twala
Research and Development Academy
Prototyping technology is becoming vital in the business as a means of cutting costs and manufacturing time. At present, reverse engineering and rapid prototyping are important technologies that enhance prototype development. The traditional approaches require various intricate processes, such as selective heat sintering (SHS), digital-light-processing printer (DLP), remote distributed rapid prototyping model (RDRP), Stereo Lithography (STL) models, and reconstructing computer-aided design (CAD) models from scanned point data and these approaches has limitations in terms of time-consuming and expert knowledge required for automation. This study aims to explore the significance of Industry 4.0 and its impact on rapid prototyping. The study also addresses rapid prototyping in computer network architecture; digital-light-processing printers (DLP) in rapid prototyping, and software-defined network (SDN) networks in the context of rapid prototyping. Along with this powder bed fusion (PBF) method and electron beam melting (EBM) are included in the manuscript. Based on our exploration, the study suggested vital recommendations for the advancement in rapid prototyping using Industry 4.0
Rajesh Singh, Anita Gehlot, Shaik Vaseem Akram, Amit Kumar Thakur, Lovi Raj Gupta, Neeraj Priyadarshi, and Bhekisipho Twala
Research and Development Academy
The potential assimilation of Industry 4.0 technologies across diverse sectors unlocks the pathways to achieve sustainability through innovative infrastructure with sustainable approaches. The World Travel & Tourism Council’s (WTTC) 2023 report emphasizes the impact of the hospitality industry contributes $9.5 trillion to the gross domestic product (GDP) and provides a workforce of 320 million globally and also strives towards meeting sustainability. Driven by the facts above, this study conducted a review to explore the potentiality of Industry 4.0 technologies specifically focused on meeting sustainability. Along with the review, the study has proposed a scalable-based architecture with the assimilation of many Industry 4.0 technologies. Further, this study has analyzed the real-world examples of Industry 4.0 technologies adoption in the hospitality industry with an objective of innovation and sustainable practices. Finally, the articles discussed the recommendations that can empower the establishment of resilient infrastructure through Industry 4.0 technologies.
Sibusiso Reuben Bakana, Yongfei Zhang, and Bhekisipho Twala
IEEE
Lightweight and efficient wildlife monitoring algorithms, including pose estimation, that can operate on edge devices with limited computational resources are urgently needed for wildlife studies and protections. To reduce computational costs while maintaining accuracy in animal pose estimation, this paper introduces WildPose, an HRNet-w32 based model, designed, which integrates Efficient Channel Attention (ECA) to enhance important feature representations without complex operations, Non-Local Blocks (NLB) to capture long-range contextual information and handle occlusions, and Depthwise Convolutions (DWConv) to reduce computational complexity and parameters. Additionally, the Online Hard Keypoint Mining with Mean Squared Error (OHKMSE) loss function is employed to improve accuracy for occluded keypoints detection by focusing on harder keypoints. When evaluated on the largest wildlife dataset of Animal Kingdom, WildPose demonstrated a nice trade-off between accuracy and efficiency, 75% reduction in parameters and 65% reduction in GFLOPs, while achieving an increased Percentage of Correct Keypoints (PCK) for occluded keypoints such as the hip, ankle, and tail. Therefore, WildPose can serve as a practical solution for real-time wildlife monitoring and artificial intelligence-based ecological studies, especially in resource-constrained environments.