{"id":{"repo_id":"bu","oai_identifier":"oai:open.bu.edu:2144/49894"},"canonical_url":"https://search.ndltd.org/etd/bu/oai:open.bu.edu:2144/49894","repository":{"repo_id":"bu","name":"Boston University","base_url":"https://open.bu.edu/oai/request"},"display":{"title":"Proposing coarse to fine grained prediction and hard negative mining for open set 3D object detection","abstract":"In recent years, there has been a remarkable advancement in robotics, autonomous vehicles, and augmented reality technologies, leading to a surge of interest and research activities in 3D learning. Among the 3D recognition research, a significant portion focuses on closed-set detection, overlooking the inherently open nature of real-world scenarios. Furthermore, the scarcity of large-scale 3D datasets, compounded by the high cost of data collection poses a substantial challenge for researchers in this domain. Motivated by these limitations, our work centers on enhancing 3D object detection within an open-set setting. Our work addresses two key research questions within the realm of open-set 3D object recognition, which has not been addressed in prior literature. Firstly, we investigate the efficacy of employing a coarse to fine-grained prediction strategy. This approach aims to enhance the performance of visually similar categories while maintaining the performance of other categories within the dataset. Secondly, we explore the utilization of offline hard negative mining, specifically targeting challenging samples within the text and image modalities to align with the point cloud encoder. This methodology leads to robust performance, particularly on syntactically similar categories. Through these approaches, our study contributes to the advancement of open-set object detection in 3D learning, thereby addressing critical gaps in current research efforts.","abstract_html":"In recent years, there has been a remarkable advancement in robotics, autonomous vehicles, and augmented reality technologies, leading to a surge of interest and research activities in 3D learning. Among the 3D recognition research, a significant portion focuses on closed-set detection, overlooking the inherently open nature of real-world scenarios. Furthermore, the scarcity of large-scale 3D datasets, compounded by the high cost of data collection poses a substantial challenge for researchers in this domain. Motivated by these limitations, our work centers on enhancing 3D object detection within an open-set setting. Our work addresses two key research questions within the realm of open-set 3D object recognition, which has not been addressed in prior literature. Firstly, we investigate the efficacy of employing a coarse to fine-grained prediction strategy. This approach aims to enhance the performance of visually similar categories while maintaining the performance of other categories within the dataset. Secondly, we explore the utilization of offline hard negative mining, specifically targeting challenging samples within the text and image modalities to align with the point cloud encoder. This methodology leads to robust performance, particularly on syntactically similar categories. Through these approaches, our study contributes to the advancement of open-set object detection in 3D learning, thereby addressing critical gaps in current research efforts.","abstract_has_math":false,"creators":["Khatri, Harsh"],"institution":"Boston University","degree_name":"Master of Science","degree_level":"masters","degree_discipline":"Computer Science","degree_department":null,"school":null,"contributors":[],"advisors":["Plummer, Bryan A."],"committee_chairs":[],"committee_members":[],"year":2024,"date_issued":"2024","date_published":"2024","updated_at":"2026-11-18T19:28:04Z","subjects":["Computer science"],"languages":["en_US"],"rights":[],"rights_urls":[],"identifier_entries":[],"person_identifier_entries":[]},"links":{"outbound_url":"https://hdl.handle.net/2144/49894","outbound_label":"Handle","outbound_source":"dc:identifier.uri","record_url":"https://hdl.handle.net/2144/49894","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://open.bu.edu/oai/request?verb=GetRecord&metadataPrefix=dim&identifier=oai%3Aopen.bu.edu%3A2144%2F49894","prefix":"dim"},"metadata_groups":[{"id":"people","label":"People","entries":[{"key":"dc:contributor.advisor","label":"Advisor","values":["Plummer, Bryan A."]},{"key":"dc:creator","label":"Author","values":["Khatri, Harsh"]}]},{"id":"academic_context","label":"Academic Context","entries":[{"key":"dc:date.accessioned","label":"Dc Date Accessioned","values":["2025-03-06T15:25:12Z"]},{"key":"dc:date.available","label":"Dc Date Available","values":["2025-03-06T15:25:12Z"]},{"key":"dc:date.issued","label":"Date","values":["2024"]},{"key":"dc:date.updated","label":"Dc Date Updated","values":["2025-03-05T02:01:24Z"]},{"key":"dc:type","label":"Dc Type","values":["Thesis/Dissertation"]},{"key":"thesis:degree_discipline","label":"Discipline","values":["Computer Science"]},{"key":"thesis:degree_level","label":"Degree Level","values":["masters"]},{"key":"thesis:degree_name","label":"Degree Name","values":["Master of Science"]},{"key":"thesis:institution_name","label":"Thesis Institution Name","values":["Boston University"]}]},{"id":"subjects_keywords","label":"Subjects and Keywords","entries":[{"key":"dc:subject","label":"Dc Subject","values":["Computer science"]}]},{"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.uri","label":"Identifier URI","values":["https://hdl.handle.net/2144/49894"]}]},{"id":"additional","label":"Additional Metadata","entries":[{"key":"dc:description","label":"Description","values":["2024"]},{"key":"dc:description.abstract","label":"Abstract","values":["In recent years, there has been a remarkable advancement in robotics, autonomous vehicles, and augmented reality technologies, leading to a surge of interest and research activities in 3D learning. Among the 3D recognition research, a significant portion focuses on closed-set detection, overlooking the inherently open nature of real-world scenarios. Furthermore, the scarcity of large-scale 3D datasets, compounded by the high cost of data collection poses a substantial challenge for researchers in this domain. Motivated by these limitations, our work centers on enhancing 3D object detection within an open-set setting. Our work addresses two key research questions within the realm of open-set 3D object recognition, which has not been addressed in prior literature. Firstly, we investigate the efficacy of employing a coarse to fine-grained prediction strategy. This approach aims to enhance the performance of visually similar categories while maintaining the performance of other categories within the dataset. Secondly, we explore the utilization of offline hard negative mining, specifically targeting challenging samples within the text and image modalities to align with the point cloud encoder. This methodology leads to robust performance, particularly on syntactically similar categories. Through these approaches, our study contributes to the advancement of open-set object detection in 3D learning, thereby addressing critical gaps in current research efforts."]},{"key":"dc:title","label":"Title","values":["Proposing coarse to fine grained prediction and hard negative mining for open set 3D object detection"]}]}],"canonical_facts":{"dc:contributor.advisor":["Plummer, Bryan A."],"dc:creator":["Khatri, Harsh"],"dc:date.accessioned":["2025-03-06T15:25:12Z"],"dc:date.available":["2025-03-06T15:25:12Z"],"dc:date.issued":["2024"],"dc:date.updated":["2025-03-05T02:01:24Z"],"dc:description":["2024"],"dc:description.abstract":["In recent years, there has been a remarkable advancement in robotics, autonomous vehicles, and augmented reality technologies, leading to a surge of interest and research activities in 3D learning. Among the 3D recognition research, a significant portion focuses on closed-set detection, overlooking the inherently open nature of real-world scenarios. Furthermore, the scarcity of large-scale 3D datasets, compounded by the high cost of data collection poses a substantial challenge for researchers in this domain. Motivated by these limitations, our work centers on enhancing 3D object detection within an open-set setting. Our work addresses two key research questions within the realm of open-set 3D object recognition, which has not been addressed in prior literature. Firstly, we investigate the efficacy of employing a coarse to fine-grained prediction strategy. This approach aims to enhance the performance of visually similar categories while maintaining the performance of other categories within the dataset. Secondly, we explore the utilization of offline hard negative mining, specifically targeting challenging samples within the text and image modalities to align with the point cloud encoder. This methodology leads to robust performance, particularly on syntactically similar categories. Through these approaches, our study contributes to the advancement of open-set object detection in 3D learning, thereby addressing critical gaps in current research efforts."],"dc:identifier.uri":["https://hdl.handle.net/2144/49894"],"dc:language.iso":["en_US"],"dc:subject":["Computer science"],"dc:title":["Proposing coarse to fine grained prediction and hard negative mining for open set 3D object detection"],"dc:type":["Thesis/Dissertation"],"thesis:degree_discipline":["Computer Science"],"thesis:degree_level":["masters"],"thesis:degree_name":["Master of Science"],"thesis:institution_name":["Boston University"]},"updated_at":"2026-11-18T19:28:04Z"}