Towards open world semi supervised detection

Traditional object detection networks work with large amounts of labeled data and under the assumption of a closed set, such that the test data only contains instances of classes already seen in the training set. These assumptions are challenged when deploying these methods in the wild. In this work we introduce Open World Semi Supervised Object Detection (OWSSD), a semi supervised learning framework that works in the open world setup. OWSSD effectively captures the novelty of unseen data compared to seen data and updates the detection framework to discover new classes on the fly.

Name thesis:degree_name

M.S.

Level thesis:degree_level

Thesis

Discipline thesis:degree_discipline

Computer Science

Grantor

University of Illinois at Urbana-Champaign

Year dc:date

2022

Author dc:creator

• Allabadi, Garvita

Contributors dc:contributor

• Adve, Vikram

• Open world
• Semi supervised learning
• Object Detection

Statement dc:rights

• Copyright 2022 Garvita Allabadi

Language dc:language

en, eng

Handle dc:identifier

https://hdl.handle.net/2142/115621

OAI record oai:identifier

https://www.ideals.illinois.edu/oai-pmh?verb=GetRecord&metadataPrefix=etdms&identifier=oai%3Awww.ideals.illinois.edu%3A2142%2F115621