{"id":{"repo_id":"cornell","oai_identifier":"oai:ecommons.cornell.edu:1813/67404"},"canonical_url":"https://search.ndltd.org/etd/cornell/oai:ecommons.cornell.edu:1813/67404","repository":{"repo_id":"cornell","name":"Cornell University","base_url":"https://ecommons.cornell.edu/server/oai/request"},"display":{"title":"Lowshot Segmentation","abstract":"Learning from limited supervision has become an area of interest in machine learning because deep learning systems have demonstrated a dependence on large, labeled data sets which can be costly and sometimes unfeasible to obtain. A subset of this endeavor is the task of low-shot learning which is learning from a handful of labeled examples and has seen recent attention in image recognition. In this work we investigate the low-shot learning paradigm in the context of semantic segmentation, the challenges that it presents, and develop techniques for addressing these challenges. We first explore multiple ways in which limited labeling can occur in the task of semantic segmentation which is not a straight forward extension of its image recognition counter-part. Next we present novel techniques for increasing performance which include a form of learned data augmentation and incorporating inductive biases about local context to leverage unlabeled data. Finally we shift the experiments to a more natural and general setting in which the data naturally follows a heavy-tail distribution over the class frequency and size rather than artificially limiting labeled data for certain classes.","abstract_html":"Learning from limited supervision has become an area of interest in machine learning because deep learning systems have demonstrated a dependence on large, labeled data sets which can be costly and sometimes unfeasible to obtain. A subset of this endeavor is the task of low-shot learning which is learning from a handful of labeled examples and has seen recent attention in image recognition. In this work we investigate the low-shot learning paradigm in the context of semantic segmentation, the challenges that it presents, and develop techniques for addressing these challenges. We first explore multiple ways in which limited labeling can occur in the task of semantic segmentation which is not a straight forward extension of its image recognition counter-part. Next we present novel techniques for increasing performance which include a form of learned data augmentation and incorporating inductive biases about local context to leverage unlabeled data. Finally we shift the experiments to a more natural and general setting in which the data naturally follows a heavy-tail distribution over the class frequency and size rather than artificially limiting labeled data for certain classes.","abstract_has_math":false,"creators":["Wallingford, Matthew C"],"institution":"Cornell University","degree_name":"M.S., Computer Science","degree_level":"Master of Science","degree_discipline":"Computer Science","degree_department":null,"school":null,"contributors":[],"advisors":[],"committee_chairs":["Hariharan, Bharath"],"committee_members":["Sridharan, Karthik"],"year":2019,"date_issued":"2019-05-30","date_published":"2019-05-30","updated_at":"2026-11-19T03:20:50Z","subjects":["Lowshot","Semantic Segmentation","computer vision","Computer science","machine learning","Fewshot","Heavy-tailed Distribution"],"languages":["en_US"],"rights":[],"rights_urls":[],"identifier_entries":[{"key":"dc:identifier.doi","label":"DOI","values":["https://doi.org/10.7298/ej6p-2g64"],"render_values":[{"text":"https://doi.org/10.7298/ej6p-2g64","href":"https://doi.org/10.7298/ej6p-2g64","code":true}]},{"key":"dc:identifier.bibid","label":"Dc Identifier Bibid","values":["11050386"],"render_values":[{"text":"11050386","href":null,"code":true}]},{"key":"dc:identifier.other","label":"Dc Identifier Other","values":["ProQuest Submission ID: 10550","ProQuest Publication ID: 13883710"],"render_values":[{"text":"ProQuest Submission ID: 10550","href":null,"code":true},{"text":"ProQuest Publication ID: 13883710","href":null,"code":true}]}],"person_identifier_entries":[]},"links":{"outbound_url":"https://hdl.handle.net/1813/67404","outbound_label":"Handle","outbound_source":"dc:identifier.uri","record_url":"https://hdl.handle.net/1813/67404","record_label":"Repository record","record_kind":"handle","record_source":"dc:identifier.uri","download_url":null,"download_label":null,"download_source":null,"access_status":null,"embargo_until":null,"access_note":null},"source_record":{"url":"https://ecommons.cornell.edu/server/oai/request?verb=GetRecord&metadataPrefix=dim&identifier=oai%3Aecommons.cornell.edu%3A1813%2F67404","prefix":"dim"},"metadata_groups":[{"id":"people","label":"People","entries":[{"key":"dc:contributor.chair","label":"Chair","values":["Hariharan, Bharath"]},{"key":"dc:contributor.committeemember","label":"Committee Member","values":["Sridharan, Karthik"]},{"key":"dc:creator","label":"Author","values":["Wallingford, Matthew C"]}]},{"id":"academic_context","label":"Academic Context","entries":[{"key":"dc:date.accessioned","label":"Dc Date Accessioned","values":["2019-10-15T15:31:14Z"]},{"key":"dc:date.available","label":"Dc Date Available","values":["2019-10-15T15:31:14Z"]},{"key":"dc:date.issued","label":"Date","values":["2019-05-30"]},{"key":"dc:type","label":"Dc Type","values":["dissertation or thesis"]},{"key":"thesis:degree_discipline","label":"Discipline","values":["Computer Science"]},{"key":"thesis:degree_level","label":"Degree Level","values":["Master of Science"]},{"key":"thesis:degree_name","label":"Degree Name","values":["M.S., Computer Science"]},{"key":"thesis:institution_name","label":"Thesis Institution Name","values":["Cornell University"]}]},{"id":"subjects_keywords","label":"Subjects and Keywords","entries":[{"key":"dc:subject","label":"Dc Subject","values":["Lowshot","Semantic Segmentation","computer vision","Computer science","machine learning","Fewshot","Heavy-tailed Distribution"]}]},{"id":"language_rights","label":"Language and Rights","entries":[{"key":"dc:language.iso","label":"Language (ISO)","values":["en_US"]}]},{"id":"identifiers","label":"Identifiers","entries":[{"key":"dc:identifier.bibid","label":"Dc Identifier Bibid","values":["11050386"]},{"key":"dc:identifier.doi","label":"DOI","values":["https://doi.org/10.7298/ej6p-2g64"]},{"key":"dc:identifier.other","label":"Dc Identifier Other","values":["ProQuest Submission ID: 10550","ProQuest Publication ID: 13883710"]},{"key":"dc:identifier.uri","label":"Identifier URI","values":["https://hdl.handle.net/1813/67404"]}]},{"id":"additional","label":"Additional Metadata","entries":[{"key":"cerif:openaire","label":"Cerif Openaire","values":["<Publication xmlns=\"https://www.openaire.eu/cerif-profile/1.1/\" id=\"4d0f1b2c-d727-4cb8-b76b-5115e7c61016\"> <Type xmlns=\"https://www.openaire.eu/cerif-profile/vocab/COAR_Publication_Types\">http://purl.org/coar/resource_type/c_1843</Type> <Language>en_US</Language> <Title>Lowshot Segmentation</Title> <PublishedIn> <Publication> </Publication> </PublishedIn> <PublicationDate>2019-05-30</PublicationDate> <DOI>https://doi.org/10.7298/ej6p-2g64</DOI> <Authors> <Author> <DisplayName>Wallingford, Matthew C</DisplayName> <Affiliation> <OrgUnit> </OrgUnit> </Affiliation> </Author> </Authors> <Editors> </Editors> <Publishers> <Publisher> <OrgUnit /> </Publisher> </Publishers> <Keyword>Lowshot</Keyword> <Keyword>Semantic Segmentation</Keyword> <Keyword>computer vision</Keyword> <Keyword>Computer science</Keyword> <Keyword>machine learning</Keyword> <Keyword>Fewshot</Keyword> <Keyword>Heavy-tailed Distribution</Keyword> <Abstract>Learning from limited supervision has become an area of interest in machine learning because deep learning systems have demonstrated a dependence on large, labeled data sets which can be costly and sometimes unfeasible to obtain. A subset of this endeavor is the task of low-shot learning which is learning from a handful of labeled examples and has seen recent attention in image recognition. In this work we investigate the low-shot learning paradigm in the context of semantic segmentation, the challenges that it presents, and develop techniques for addressing these challenges. We first explore multiple ways in which limited labeling can occur in the task of semantic segmentation which is not a straight forward extension of its image recognition counter-part. Next we present novel techniques for increasing performance which include a form of learned data augmentation and incorporating inductive biases about local context to leverage unlabeled data. Finally we shift the experiments to a more natural and general setting in which the data naturally follows a heavy-tail distribution over the class frequency and size rather than artificially limiting labeled data for certain classes.</Abstract> <Access xmlns=\"http://purl.org/coar/access_right\" > </Access> </Publication>"]},{"key":"cris:virtual.author","label":"Cris Virtual Author","values":["Wallingford, Matthew C"]},{"key":"cris:virtualsource.collection","label":"Cris Virtualsource Collection","values":["5893a6ea-7af3-41d7-abc6-04bcd26ab5df"]},{"key":"dc:description.abstract","label":"Abstract","values":["Learning from limited supervision has become an area of interest in machine learning because deep learning systems have demonstrated a dependence on large, labeled data sets which can be costly and sometimes unfeasible to obtain. A subset of this endeavor is the task of low-shot learning which is learning from a handful of labeled examples and has seen recent attention in image recognition. In this work we investigate the low-shot learning paradigm in the context of semantic segmentation, the challenges that it presents, and develop techniques for addressing these challenges. We first explore multiple ways in which limited labeling can occur in the task of semantic segmentation which is not a straight forward extension of its image recognition counter-part. Next we present novel techniques for increasing performance which include a form of learned data augmentation and incorporating inductive biases about local context to leverage unlabeled data. Finally we shift the experiments to a more natural and general setting in which the data naturally follows a heavy-tail distribution over the class frequency and size rather than artificially limiting labeled data for certain classes."]},{"key":"dc:description.collection","label":"Dc Description Collection","values":["Cornell Theses and Dissertations"]},{"key":"dc:format.mimetype","label":"Dc Format Mimetype","values":["application/pdf"]},{"key":"dcterms:license","label":"Dcterms License","values":["https://hdl.handle.net/1813/59810"]},{"key":"dc:title","label":"Title","values":["Lowshot Segmentation"]},{"key":"dspace:entity.type","label":"Dspace Entity Type","values":["Publication"]},{"key":"others:access_status","label":"Others Access Status","values":["open.access"]}]}],"canonical_facts":{"cerif:openaire":["<Publication xmlns=\"https://www.openaire.eu/cerif-profile/1.1/\" id=\"4d0f1b2c-d727-4cb8-b76b-5115e7c61016\"> <Type xmlns=\"https://www.openaire.eu/cerif-profile/vocab/COAR_Publication_Types\">http://purl.org/coar/resource_type/c_1843</Type> <Language>en_US</Language> <Title>Lowshot Segmentation</Title> <PublishedIn> <Publication> </Publication> </PublishedIn> <PublicationDate>2019-05-30</PublicationDate> <DOI>https://doi.org/10.7298/ej6p-2g64</DOI> <Authors> <Author> <DisplayName>Wallingford, Matthew C</DisplayName> <Affiliation> <OrgUnit> </OrgUnit> </Affiliation> </Author> </Authors> <Editors> </Editors> <Publishers> <Publisher> <OrgUnit /> </Publisher> </Publishers> <Keyword>Lowshot</Keyword> <Keyword>Semantic Segmentation</Keyword> <Keyword>computer vision</Keyword> <Keyword>Computer science</Keyword> <Keyword>machine learning</Keyword> <Keyword>Fewshot</Keyword> <Keyword>Heavy-tailed Distribution</Keyword> <Abstract>Learning from limited supervision has become an area of interest in machine learning because deep learning systems have demonstrated a dependence on large, labeled data sets which can be costly and sometimes unfeasible to obtain. A subset of this endeavor is the task of low-shot learning which is learning from a handful of labeled examples and has seen recent attention in image recognition. In this work we investigate the low-shot learning paradigm in the context of semantic segmentation, the challenges that it presents, and develop techniques for addressing these challenges. We first explore multiple ways in which limited labeling can occur in the task of semantic segmentation which is not a straight forward extension of its image recognition counter-part. Next we present novel techniques for increasing performance which include a form of learned data augmentation and incorporating inductive biases about local context to leverage unlabeled data. Finally we shift the experiments to a more natural and general setting in which the data naturally follows a heavy-tail distribution over the class frequency and size rather than artificially limiting labeled data for certain classes.</Abstract> <Access xmlns=\"http://purl.org/coar/access_right\" > </Access> </Publication>"],"cris:virtual.author":["Wallingford, Matthew C"],"cris:virtualsource.collection":["5893a6ea-7af3-41d7-abc6-04bcd26ab5df"],"dc:contributor.chair":["Hariharan, Bharath"],"dc:contributor.committeemember":["Sridharan, Karthik"],"dc:creator":["Wallingford, Matthew C"],"dc:date.accessioned":["2019-10-15T15:31:14Z"],"dc:date.available":["2019-10-15T15:31:14Z"],"dc:date.issued":["2019-05-30"],"dc:description.abstract":["Learning from limited supervision has become an area of interest in machine learning because deep learning systems have demonstrated a dependence on large, labeled data sets which can be costly and sometimes unfeasible to obtain. A subset of this endeavor is the task of low-shot learning which is learning from a handful of labeled examples and has seen recent attention in image recognition. In this work we investigate the low-shot learning paradigm in the context of semantic segmentation, the challenges that it presents, and develop techniques for addressing these challenges. We first explore multiple ways in which limited labeling can occur in the task of semantic segmentation which is not a straight forward extension of its image recognition counter-part. Next we present novel techniques for increasing performance which include a form of learned data augmentation and incorporating inductive biases about local context to leverage unlabeled data. Finally we shift the experiments to a more natural and general setting in which the data naturally follows a heavy-tail distribution over the class frequency and size rather than artificially limiting labeled data for certain classes."],"dc:description.collection":["Cornell Theses and Dissertations"],"dc:format.mimetype":["application/pdf"],"dc:identifier.bibid":["11050386"],"dc:identifier.doi":["https://doi.org/10.7298/ej6p-2g64"],"dc:identifier.other":["ProQuest Submission ID: 10550","ProQuest Publication ID: 13883710"],"dc:identifier.uri":["https://hdl.handle.net/1813/67404"],"dc:language.iso":["en_US"],"dc:subject":["Lowshot","Semantic Segmentation","computer vision","Computer science","machine learning","Fewshot","Heavy-tailed Distribution"],"dcterms:license":["https://hdl.handle.net/1813/59810"],"dc:title":["Lowshot Segmentation"],"dc:type":["dissertation or thesis"],"dspace:entity.type":["Publication"],"others:access_status":["open.access"],"thesis:degree_discipline":["Computer Science"],"thesis:degree_level":["Master of Science"],"thesis:degree_name":["M.S., Computer Science"],"thesis:institution_name":["Cornell University"]},"updated_at":"2026-11-19T03:20:50Z"}