{"id":{"repo_id":"uiuc","oai_identifier":"oai:www.ideals.illinois.edu:2142/105809"},"canonical_url":"https://search.dev.ndltd.org/etd/uiuc/oai:www.ideals.illinois.edu:2142/105809","repository":{"repo_id":"uiuc","name":"University of Illinois - Urbana-Champaign","base_url":"https://www.ideals.illinois.edu/oai-pmh"},"display":{"title":"Data generalization for new classes with a single instance via automatic style labeling and transfer","abstract":"In order to synthesize new images from a specific class, most generative models like Generative Adversarial Nets (GANs) require a large amount of data from this class. In other words, modern generative models often lack the ability to create new samples belonging to an unseen class from which they have observed only one instance. In this thesis, we propose a model that can generalize a single instance from an unseen class and create a whole data distribution of the class by learning how data from other classes vary within their own distributions, and transferring this information to the new class. We show that the new samples generated by our model not only preserve the essential visual features for them to be recognized as in the same class that the source instance is from, but also exhibit variety. Experiments on the MNIST dataset show that after hiding away one class of digits and training only on the data of the remaining nine classes, our model can successfully generate new images of the hidden class with controllable features, given just a single image from that class.","abstract_html":"In order to synthesize new images from a specific class, most generative models like Generative Adversarial Nets (GANs) require a large amount of data from this class. In other words, modern generative models often lack the ability to create new samples belonging to an unseen class from which they have observed only one instance. In this thesis, we propose a model that can generalize a single instance from an unseen class and create a whole data distribution of the class by learning how data from other classes vary within their own distributions, and transferring this information to the new class. We show that the new samples generated by our model not only preserve the essential visual features for them to be recognized as in the same class that the source instance is from, but also exhibit variety. Experiments on the MNIST dataset show that after hiding away one class of digits and training only on the data of the remaining nine classes, our model can successfully generate new images of the hidden class with controllable features, given just a single image from that class.","abstract_has_math":false,"creators":["Zou, Yuxuan"],"institution":"University of Illinois at Urbana-Champaign","degree_name":"M.S.","degree_level":"Thesis","degree_discipline":"Computer Science","degree_department":null,"school":null,"contributors":["Koyejo, Sanmi"],"advisors":[],"committee_chairs":[],"committee_members":[],"year":2019,"date_issued":"2019-11-26T20:49:26Z","date_published":"2019-11-26T20:49:26Z","updated_at":"2026-07-22T22:24:45Z","subjects":["Style Transfer","Data Generation"],"languages":["en"],"rights":["Copyright 2019 Yuxuan Zou"],"rights_urls":[],"identifier_entries":[]},"links":{"outbound_url":"http://hdl.handle.net/2142/105809","outbound_label":"Handle","outbound_source":"dc:identifier"},"metadata_groups":[{"id":"people","label":"People","entries":[{"key":"dc:contributor","label":"Contributor","values":["Koyejo, Sanmi"]},{"key":"dc:creator","label":"Author","values":["Zou, Yuxuan"]}]},{"id":"academic_context","label":"Academic Context","entries":[{"key":"dc:date","label":"Dc Date","values":["2019-11-26T20:49:26Z","2021-11-27T10:15:26Z","2019-07-11","2019-08"]},{"key":"dc:type","label":"Dc Type","values":["text"]},{"key":"thesis:degree_discipline","label":"Discipline","values":["Computer Science"]},{"key":"thesis:degree_level","label":"Degree Level","values":["Thesis"]},{"key":"thesis:degree_name","label":"Degree Name","values":["M.S."]},{"key":"thesis:institution_name","label":"Thesis Institution Name","values":["University of Illinois at Urbana-Champaign"]}]},{"id":"subjects_keywords","label":"Subjects and Keywords","entries":[{"key":"dc:subject","label":"Dc Subject","values":["Style Transfer","Data Generation"]}]},{"id":"language_rights","label":"Language and Rights","entries":[{"key":"dc:language","label":"Dc Language","values":["en"]},{"key":"dc:rights","label":"Dc Rights","values":["Copyright 2019 Yuxuan Zou"]}]},{"id":"identifiers","label":"Identifiers","entries":[{"key":"dc:identifier","label":"Identifier","values":["http://hdl.handle.net/2142/105809"]}]},{"id":"additional","label":"Additional Metadata","entries":[{"key":"dc:description","label":"Description","values":["In order to synthesize new images from a specific class, most generative models like Generative Adversarial Nets (GANs) require a large amount of data from this class. In other words, modern generative models often lack the ability to create new samples belonging to an unseen class from which they have observed only one instance. In this thesis, we propose a model that can generalize a single instance from an unseen class and create a whole data distribution of the class by learning how data from other classes vary within their own distributions, and transferring this information to the new class. We show that the new samples generated by our model not only preserve the essential visual features for them to be recognized as in the same class that the source instance is from, but also exhibit variety. Experiments on the MNIST dataset show that after hiding away one class of digits and training only on the data of the remaining nine classes, our model can successfully generate new images of the hidden class with controllable features, given just a single image from that class.","Submission published under a 24 month embargo labeled 'U of I Access', the embargo will last until 2021-08-01","The student, Yuxuan Zou, accepted the attached license on 2019-07-10 at 16:48.","The student, Yuxuan Zou, submitted this Thesis for approval on 2019-07-10 at 16:54.","This Thesis was approved for publication on 2019-07-11 at 12:54.","DSpace SAF Submission Ingestion Package generated from Vireo submission #14250 on 2019-11-26 at 13:05:19","Made available in DSpace on 2019-11-26T20:49:26Z (GMT). No. of bitstreams: 2 ZOU-THESIS-2019.pdf: 1074175 bytes, checksum: d82e251b61c63f0f3bb6e90c67426092 (MD5) LICENSE.txt: 4207 bytes, checksum: e213044c65083c4559c71b75f5cdcd7e (MD5) Previous issue date: 2019-07-11","Embargo set by: Seth Robbins for item 112954 Lift date: 2021-11-26T20:49:41Z Reason: Author requested U of Illinois access only (OA after 2yrs) in Vireo ETD system","U of I Only Restriction Lifted for Item 112954 on 2021-11-27T10:15:26Z."]},{"key":"dc:format","label":"Dc Format","values":["application/pdf"]},{"key":"dc:title","label":"Title","values":["Data generalization for new classes with a single instance via automatic style labeling and transfer"]}]}],"canonical_facts":{"dc:contributor":["Koyejo, Sanmi"],"dc:creator":["Zou, Yuxuan"],"dc:date":["2019-11-26T20:49:26Z","2021-11-27T10:15:26Z","2019-07-11","2019-08"],"dc:description":["In order to synthesize new images from a specific class, most generative models like Generative Adversarial Nets (GANs) require a large amount of data from this class. In other words, modern generative models often lack the ability to create new samples belonging to an unseen class from which they have observed only one instance. In this thesis, we propose a model that can generalize a single instance from an unseen class and create a whole data distribution of the class by learning how data from other classes vary within their own distributions, and transferring this information to the new class. We show that the new samples generated by our model not only preserve the essential visual features for them to be recognized as in the same class that the source instance is from, but also exhibit variety. Experiments on the MNIST dataset show that after hiding away one class of digits and training only on the data of the remaining nine classes, our model can successfully generate new images of the hidden class with controllable features, given just a single image from that class.","Submission published under a 24 month embargo labeled 'U of I Access', the embargo will last until 2021-08-01","The student, Yuxuan Zou, accepted the attached license on 2019-07-10 at 16:48.","The student, Yuxuan Zou, submitted this Thesis for approval on 2019-07-10 at 16:54.","This Thesis was approved for publication on 2019-07-11 at 12:54.","DSpace SAF Submission Ingestion Package generated from Vireo submission #14250 on 2019-11-26 at 13:05:19","Made available in DSpace on 2019-11-26T20:49:26Z (GMT). No. of bitstreams: 2 ZOU-THESIS-2019.pdf: 1074175 bytes, checksum: d82e251b61c63f0f3bb6e90c67426092 (MD5) LICENSE.txt: 4207 bytes, checksum: e213044c65083c4559c71b75f5cdcd7e (MD5) Previous issue date: 2019-07-11","Embargo set by: Seth Robbins for item 112954 Lift date: 2021-11-26T20:49:41Z Reason: Author requested U of Illinois access only (OA after 2yrs) in Vireo ETD system","U of I Only Restriction Lifted for Item 112954 on 2021-11-27T10:15:26Z."],"dc:format":["application/pdf"],"dc:identifier":["http://hdl.handle.net/2142/105809"],"dc:language":["en"],"dc:rights":["Copyright 2019 Yuxuan Zou"],"dc:subject":["Style Transfer","Data Generation"],"dc:title":["Data generalization for new classes with a single instance via automatic style labeling and transfer"],"dc:type":["text"],"thesis:degree_discipline":["Computer Science"],"thesis:degree_level":["Thesis"],"thesis:degree_name":["M.S."],"thesis:institution_name":["University of Illinois at Urbana-Champaign"]},"updated_at":"2026-07-22T22:24:45Z"}