{"id":{"repo_id":"uwtsd","oai_identifier":"oai:repository.uwtsd.ac.uk:2062"},"canonical_url":"https://search.dev.ndltd.org/etd/uwtsd/oai:repository.uwtsd.ac.uk:2062","repository":{"repo_id":"uwtsd","name":"University of Wales Trinity Saint David","base_url":"https://repository.uwtsd.ac.uk/cgi/oai2"},"display":{"title":"Fast head profile estimation using curvature, derivatives and deep learning methods","abstract":"Fast estimation of head profile and posture has applications across many disciplines, for example, it can be used in sleep apnoea screening and orthodontic examination or could support a suitable physiotherapy regime. Consequently, this thesis focuses on the investigation of methods to estimate head profile and posture efficiently and accurately, and results in the development and evaluation of datasets, features and deep learning models that can achieve this. Accordingly, this thesis initially investigated properties of contour curves that could act as effective features to train machine learning models. Features based on curvature and the first and second Gaussian derivatives were evaluated. These outperformed established features used in the literature to train a long short-term memory recurrent neural network and produced a significant speedup in execution time where pre-filtering of a sampled dataset was required. Following on from this, a new dataset of head profile contours was generated and annotated with anthropometric cranio-facial landmarks, and a novel method of automatically improving the accuracy of the landmark positions was developed using ideas based on the curvature of a plane curve. The features identified here were extracted from the new head profile contour dataset and used to train long short-term recurrent neural networks. The best network, using Gaussian derivatives features achieved an accuracy of 91% and macro F1 score of 91%, an improvement of 51% and 71% respectively when compared with the un-processed contour feature. When using Gaussian derivative features, the network was able to regress landmarks accurately with mean absolute errors ranging from 0 to 5.3 pixels and standard deviations ranging from 0 to 6.9, respectively. End-to-end machine learning approaches, where a deep neural network learns the best features to use from the raw input data, were also investigated. Such an approach, using a one-dimensional temporal convolutional network was able to match previous classifiers in terms of accuracy and macro F1 score, and showed comparable regression abilities. However, this was at the expense of increased training times and increased inference times. This network was an order of magnitude slower when classifying and regressing contours.","abstract_html":"Fast estimation of head profile and posture has applications across many disciplines, for example, it can be used in sleep apnoea screening and orthodontic examination or could support a suitable physiotherapy regime. Consequently, this thesis focuses on the investigation of methods to estimate head profile and posture efficiently and accurately, and results in the development and evaluation of datasets, features and deep learning models that can achieve this. Accordingly, this thesis initially investigated properties of contour curves that could act as effective features to train machine learning models. Features based on curvature and the first and second Gaussian derivatives were evaluated. These outperformed established features used in the literature to train a long short-term memory recurrent neural network and produced a significant speedup in execution time where pre-filtering of a sampled dataset was required. Following on from this, a new dataset of head profile contours was generated and annotated with anthropometric cranio-facial landmarks, and a novel method of automatically improving the accuracy of the landmark positions was developed using ideas based on the curvature of a plane curve. The features identified here were extracted from the new head profile contour dataset and used to train long short-term recurrent neural networks. The best network, using Gaussian derivatives features achieved an accuracy of 91% and macro F1 score of 91%, an improvement of 51% and 71% respectively when compared with the un-processed contour feature. When using Gaussian derivative features, the network was able to regress landmarks accurately with mean absolute errors ranging from 0 to 5.3 pixels and standard deviations ranging from 0 to 6.9, respectively. End-to-end machine learning approaches, where a deep neural network learns the best features to use from the raw input data, were also investigated. Such an approach, using a one-dimensional temporal convolutional network was able to match previous classifiers in terms of accuracy and macro F1 score, and showed comparable regression abilities. However, this was at the expense of increased training times and increased inference times. This network was an order of magnitude slower when classifying and regressing contours.","abstract_has_math":false,"creators":["Dickers, Gordon"],"institution":"University of Wales Trinity Saint David","degree_name":"phd","degree_level":"doctoral","degree_discipline":null,"degree_department":null,"school":null,"contributors":[],"advisors":[],"committee_chairs":[],"committee_members":[],"year":2022,"date_issued":"2022-08","date_published":"2022-08","updated_at":"2026-07-24T05:52:56Z","subjects":["QA Mathemateg","QA75 Cyfrifiaduron electronig"],"languages":[],"rights":[],"rights_urls":[],"identifier_entries":[{"key":"dc:identifier.grantnumber","label":"Dc Identifier Grantnumber","values":["UWTSD"],"render_values":[{"text":"UWTSD","href":null,"code":true}]}]},"links":{"outbound_url":"https://doi.org/10.82227/repository.uwtsd.ac.uk.00002062","outbound_label":"DOI","outbound_source":"dc:identifier.doi"},"metadata_groups":[{"id":"people","label":"People","entries":[{"key":"dc:contributor.sponsor","label":"Sponsor","values":["University of Wales Trinity Saint David"]},{"key":"dc:creator","label":"Author","values":["Dickers, Gordon"]}]},{"id":"academic_context","label":"Academic Context","entries":[{"key":"dc:date","label":"Dc Date","values":["2022-08-09"]},{"key":"dc:date.issued","label":"Date","values":["2022-08"]},{"key":"dc:publisher.commercial","label":"Dc Publisher Commercial","values":["University of Wales Trinity Saint David"]},{"key":"dc:publisher.department","label":"Dc Publisher Department","values":["Traethodau PhD","Wales Institute for Science and Art: Applied Computing"]},{"key":"dc:publisher.institution","label":"Dc Publisher Institution","values":["University of Wales Trinity Saint David"]},{"key":"dc:relation.isreferencedby","label":"Dc Relation Isreferencedby","values":["https://repository.uwtsd.ac.uk/id/eprint/2062/"]},{"key":"dc:type","label":"Dc Type","values":["Gosodiad"]},{"key":"dc:type.qualificationlevel","label":"Dc Type Qualificationlevel","values":["doctoral"]},{"key":"dc:type.qualificationname","label":"Dc Type Qualificationname","values":["phd"]}]},{"id":"subjects_keywords","label":"Subjects and Keywords","entries":[{"key":"dc:subject","label":"Dc Subject","values":["QA Mathemateg","QA75 Cyfrifiaduron electronig"]}]},{"id":"identifiers","label":"Identifiers","entries":[{"key":"dc:identifier.doi","label":"DOI","values":["10.82227/repository.uwtsd.ac.uk.00002062"]},{"key":"dc:identifier.grantnumber","label":"Dc Identifier Grantnumber","values":["UWTSD"]},{"key":"dc:identifier.uri","label":"Identifier URI","values":["https://repository.uwtsd.ac.uk/id/eprint/2062/1/Dickers%20Gordon%20%282022%29%20Fast_profile_estimation_using_curvature_derivatives_and_deep_learning_methods.pdf"]}]},{"id":"additional","label":"Additional Metadata","entries":[{"key":"dc:description.abstract","label":"Abstract","values":["Fast estimation of head profile and posture has applications across many disciplines, for example, it can be used in sleep apnoea screening and orthodontic examination or could support a suitable physiotherapy regime. Consequently, this thesis focuses on the investigation of methods to estimate head profile and posture efficiently and accurately, and results in the development and evaluation of datasets, features and deep learning models that can achieve this. Accordingly, this thesis initially investigated properties of contour curves that could act as effective features to train machine learning models. Features based on curvature and the first and second Gaussian derivatives were evaluated. These outperformed established features used in the literature to train a long short-term memory recurrent neural network and produced a significant speedup in execution time where pre-filtering of a sampled dataset was required. Following on from this, a new dataset of head profile contours was generated and annotated with anthropometric cranio-facial landmarks, and a novel method of automatically improving the accuracy of the landmark positions was developed using ideas based on the curvature of a plane curve. The features identified here were extracted from the new head profile contour dataset and used to train long short-term recurrent neural networks. The best network, using Gaussian derivatives features achieved an accuracy of 91% and macro F1 score of 91%, an improvement of 51% and 71% respectively when compared with the un-processed contour feature. When using Gaussian derivative features, the network was able to regress landmarks accurately with mean absolute errors ranging from 0 to 5.3 pixels and standard deviations ranging from 0 to 6.9, respectively. End-to-end machine learning approaches, where a deep neural network learns the best features to use from the raw input data, were also investigated. Such an approach, using a one-dimensional temporal convolutional network was able to match previous classifiers in terms of accuracy and macro F1 score, and showed comparable regression abilities. However, this was at the expense of increased training times and increased inference times. This network was an order of magnitude slower when classifying and regressing contours."]},{"key":"dc:format","label":"Dc Format","values":["text"]},{"key":"dc:title","label":"Title","values":["Fast head profile estimation using curvature, derivatives and deep learning methods"]}]}],"canonical_facts":{"dc:contributor.sponsor":["University of Wales Trinity Saint David"],"dc:creator":["Dickers, Gordon"],"dc:date":["2022-08-09"],"dc:date.issued":["2022-08"],"dc:description.abstract":["Fast estimation of head profile and posture has applications across many disciplines, for example, it can be used in sleep apnoea screening and orthodontic examination or could support a suitable physiotherapy regime. Consequently, this thesis focuses on the investigation of methods to estimate head profile and posture efficiently and accurately, and results in the development and evaluation of datasets, features and deep learning models that can achieve this. Accordingly, this thesis initially investigated properties of contour curves that could act as effective features to train machine learning models. Features based on curvature and the first and second Gaussian derivatives were evaluated. These outperformed established features used in the literature to train a long short-term memory recurrent neural network and produced a significant speedup in execution time where pre-filtering of a sampled dataset was required. Following on from this, a new dataset of head profile contours was generated and annotated with anthropometric cranio-facial landmarks, and a novel method of automatically improving the accuracy of the landmark positions was developed using ideas based on the curvature of a plane curve. The features identified here were extracted from the new head profile contour dataset and used to train long short-term recurrent neural networks. The best network, using Gaussian derivatives features achieved an accuracy of 91% and macro F1 score of 91%, an improvement of 51% and 71% respectively when compared with the un-processed contour feature. When using Gaussian derivative features, the network was able to regress landmarks accurately with mean absolute errors ranging from 0 to 5.3 pixels and standard deviations ranging from 0 to 6.9, respectively. End-to-end machine learning approaches, where a deep neural network learns the best features to use from the raw input data, were also investigated. Such an approach, using a one-dimensional temporal convolutional network was able to match previous classifiers in terms of accuracy and macro F1 score, and showed comparable regression abilities. However, this was at the expense of increased training times and increased inference times. This network was an order of magnitude slower when classifying and regressing contours."],"dc:format":["text"],"dc:identifier.doi":["10.82227/repository.uwtsd.ac.uk.00002062"],"dc:identifier.grantnumber":["UWTSD"],"dc:identifier.uri":["https://repository.uwtsd.ac.uk/id/eprint/2062/1/Dickers%20Gordon%20%282022%29%20Fast_profile_estimation_using_curvature_derivatives_and_deep_learning_methods.pdf"],"dc:publisher.commercial":["University of Wales Trinity Saint David"],"dc:publisher.department":["Traethodau PhD","Wales Institute for Science and Art: Applied Computing"],"dc:publisher.institution":["University of Wales Trinity Saint David"],"dc:relation.isreferencedby":["https://repository.uwtsd.ac.uk/id/eprint/2062/"],"dc:subject":["QA Mathemateg","QA75 Cyfrifiaduron electronig"],"dc:title":["Fast head profile estimation using curvature, derivatives and deep learning methods"],"dc:type":["Gosodiad"],"dc:type.qualificationlevel":["doctoral"],"dc:type.qualificationname":["phd"]},"updated_at":"2026-07-24T05:52:56Z"}