{"id":{"repo_id":"rgu","oai_identifier":"oai:rgu-repository.worktribe.com:2795656"},"canonical_url":"https://search.dev.ndltd.org/etd/rgu/oai:rgu-repository.worktribe.com:2795656","repository":{"repo_id":"rgu","name":"Robert Gordon University","base_url":"https://rgu-repository.worktribe.com/oaiprovider"},"display":{"title":"Deep learning for digitising complex engineering drawings.","abstract":"Vast amounts of documents are still commonly stored in undigitised formats. Consequently, the data they contain cannot be used to its full potential, as substantial manual effort is required to analyse it. Amongst these documents, engineering drawings are considered one of the most challenging to digitise. The task involves automatically recognising all drawing components which are the symbols, text and connections. Although there has been significant improvement in computer vision due to the development of deep learning, the same progress has not been seen for engineering drawing digitisation. Most of these methods were based on traditional approaches which require manual feature selection and heuristics. This thesis presents a deep learning framework for the challenging problem of digitising complex engineering drawings. This is a fully automated approach for the processing and analysis of these drawings. It contains a set of deep learning methods for digitising the different drawing components. New methods were presented to recognise engineering symbols. Text digitisation methods were also developed. It should be noted that this represents a substantially more challenging problem compared to text digitisation in typical documents, due to reasons such as the varying text locations, orientations, and text strings often being composed of codes instead of known words. The thesis has solved inherent challenges in the field of engineering drawing digitisation. Furthermore, the thesis has opened up a new direction towards addressing the data annotation problem, by using few-shot learning for symbol detection. All of the methods presented here have been thoroughly tested on real world complex engineering drawings from different domains. These were Piping and Instrumentation diagrams from the oil and gas industry, and multiple engineering drawing types from the construction industry.","abstract_html":"Vast amounts of documents are still commonly stored in undigitised formats. Consequently, the data they contain cannot be used to its full potential, as substantial manual effort is required to analyse it. Amongst these documents, engineering drawings are considered one of the most challenging to digitise. The task involves automatically recognising all drawing components which are the symbols, text and connections. Although there has been significant improvement in computer vision due to the development of deep learning, the same progress has not been seen for engineering drawing digitisation. Most of these methods were based on traditional approaches which require manual feature selection and heuristics. This thesis presents a deep learning framework for the challenging problem of digitising complex engineering drawings. This is a fully automated approach for the processing and analysis of these drawings. It contains a set of deep learning methods for digitising the different drawing components. New methods were presented to recognise engineering symbols. Text digitisation methods were also developed. It should be noted that this represents a substantially more challenging problem compared to text digitisation in typical documents, due to reasons such as the varying text locations, orientations, and text strings often being composed of codes instead of known words. The thesis has solved inherent challenges in the field of engineering drawing digitisation. Furthermore, the thesis has opened up a new direction towards addressing the data annotation problem, by using few-shot learning for symbol detection. All of the methods presented here have been thoroughly tested on real world complex engineering drawings from different domains. These were Piping and Instrumentation diagrams from the oil and gas industry, and multiple engineering drawing types from the construction industry.","abstract_has_math":false,"creators":["Jamieson, Laura"],"institution":"Robert Gordon University","degree_name":null,"degree_level":null,"degree_discipline":null,"degree_department":null,"school":null,"contributors":[],"advisors":["E. Eyad and C. Moreno-Garcia"],"committee_chairs":[],"committee_members":[],"year":2024,"date_issued":"2024","date_published":"2024","updated_at":"2026-07-24T04:10:09Z","subjects":["Digitisation","Engineering drawings","Technical drawings","Symbol recognition","Text recognition","Deep learning","Few-shot learning","Multiclass imbalanced classification"],"languages":["en"],"rights":[],"rights_urls":[],"identifier_entries":[{"key":"dc:identifier","label":"Identifier","values":["oai:rgu-repository.worktribe.com:2795656","https://doi.org/10.48526/rgu-wt-2795656"],"render_values":[{"text":"oai:rgu-repository.worktribe.com:2795656","href":null,"code":true},{"text":"https://doi.org/10.48526/rgu-wt-2795656","href":"https://doi.org/10.48526/rgu-wt-2795656","code":true}]}]},"links":{"outbound_url":"https://rgu-repository.worktribe.com/2795656/1/JAMIESON%202024%20Deep%20learning%20for%20digitising","outbound_label":"Repository record","outbound_source":"dc:identifier.uri"},"metadata_groups":[{"id":"people","label":"People","entries":[{"key":"dc:contributor.advisor","label":"Advisor","values":["E. Eyad and C. Moreno-Garcia"]},{"key":"dc:contributor.sponsor","label":"Sponsor","values":["No Funder Acknowledged (Outputs)"]},{"key":"dc:creator","label":"Author","values":["Jamieson, Laura"]}]},{"id":"academic_context","label":"Academic Context","entries":[{"key":"dc:date","label":"Dc Date","values":["2024-10-31"]},{"key":"dc:date.issued","label":"Date","values":["2024"]},{"key":"dc:publisher.institution","label":"Dc Publisher Institution","values":["Robert Gordon University"]},{"key":"dc:relation.isreferencedby","label":"Dc Relation Isreferencedby","values":["https://rgu-repository.worktribe.com/output/2795656"]},{"key":"dc:type","label":"Dc Type","values":["Thesis"]}]},{"id":"subjects_keywords","label":"Subjects and Keywords","entries":[{"key":"dc:subject","label":"Dc Subject","values":["Digitisation","Engineering drawings","Technical drawings","Symbol recognition","Text recognition","Deep learning","Few-shot learning","Multiclass imbalanced classification"]}]},{"id":"language_rights","label":"Language and Rights","entries":[{"key":"dc:language","label":"Dc Language","values":["en"]}]},{"id":"identifiers","label":"Identifiers","entries":[{"key":"dc:identifier","label":"Identifier","values":["oai:rgu-repository.worktribe.com:2795656","https://doi.org/10.48526/rgu-wt-2795656"]},{"key":"dc:identifier.uri","label":"Identifier URI","values":["https://rgu-repository.worktribe.com/2795656/1/JAMIESON%202024%20Deep%20learning%20for%20digitising"]}]},{"id":"additional","label":"Additional Metadata","entries":[{"key":"dc:description.abstract","label":"Abstract","values":["Vast amounts of documents are still commonly stored in undigitised formats. Consequently, the data they contain cannot be used to its full potential, as substantial manual effort is required to analyse it. Amongst these documents, engineering drawings are considered one of the most challenging to digitise. The task involves automatically recognising all drawing components which are the symbols, text and connections. Although there has been significant improvement in computer vision due to the development of deep learning, the same progress has not been seen for engineering drawing digitisation. Most of these methods were based on traditional approaches which require manual feature selection and heuristics. This thesis presents a deep learning framework for the challenging problem of digitising complex engineering drawings. This is a fully automated approach for the processing and analysis of these drawings. It contains a set of deep learning methods for digitising the different drawing components. New methods were presented to recognise engineering symbols. Text digitisation methods were also developed. It should be noted that this represents a substantially more challenging problem compared to text digitisation in typical documents, due to reasons such as the varying text locations, orientations, and text strings often being composed of codes instead of known words. The thesis has solved inherent challenges in the field of engineering drawing digitisation. Furthermore, the thesis has opened up a new direction towards addressing the data annotation problem, by using few-shot learning for symbol detection. All of the methods presented here have been thoroughly tested on real world complex engineering drawings from different domains. These were Piping and Instrumentation diagrams from the oil and gas industry, and multiple engineering drawing types from the construction industry."]},{"key":"dc:title","label":"Title","values":["Deep learning for digitising complex engineering drawings."]}]}],"canonical_facts":{"dc:contributor.advisor":["E. Eyad and C. Moreno-Garcia"],"dc:contributor.sponsor":["No Funder Acknowledged (Outputs)"],"dc:creator":["Jamieson, Laura"],"dc:date":["2024-10-31"],"dc:date.issued":["2024"],"dc:description.abstract":["Vast amounts of documents are still commonly stored in undigitised formats. Consequently, the data they contain cannot be used to its full potential, as substantial manual effort is required to analyse it. Amongst these documents, engineering drawings are considered one of the most challenging to digitise. The task involves automatically recognising all drawing components which are the symbols, text and connections. Although there has been significant improvement in computer vision due to the development of deep learning, the same progress has not been seen for engineering drawing digitisation. Most of these methods were based on traditional approaches which require manual feature selection and heuristics. This thesis presents a deep learning framework for the challenging problem of digitising complex engineering drawings. This is a fully automated approach for the processing and analysis of these drawings. It contains a set of deep learning methods for digitising the different drawing components. New methods were presented to recognise engineering symbols. Text digitisation methods were also developed. It should be noted that this represents a substantially more challenging problem compared to text digitisation in typical documents, due to reasons such as the varying text locations, orientations, and text strings often being composed of codes instead of known words. The thesis has solved inherent challenges in the field of engineering drawing digitisation. Furthermore, the thesis has opened up a new direction towards addressing the data annotation problem, by using few-shot learning for symbol detection. All of the methods presented here have been thoroughly tested on real world complex engineering drawings from different domains. These were Piping and Instrumentation diagrams from the oil and gas industry, and multiple engineering drawing types from the construction industry."],"dc:identifier":["oai:rgu-repository.worktribe.com:2795656","https://doi.org/10.48526/rgu-wt-2795656"],"dc:identifier.uri":["https://rgu-repository.worktribe.com/2795656/1/JAMIESON%202024%20Deep%20learning%20for%20digitising"],"dc:language":["en"],"dc:publisher.institution":["Robert Gordon University"],"dc:relation.isreferencedby":["https://rgu-repository.worktribe.com/output/2795656"],"dc:subject":["Digitisation","Engineering drawings","Technical drawings","Symbol recognition","Text recognition","Deep learning","Few-shot learning","Multiclass imbalanced classification"],"dc:title":["Deep learning for digitising complex engineering drawings."],"dc:type":["Thesis"]},"updated_at":"2026-07-24T04:10:09Z"}