Online active proposal set generation for weakly supervised object detection
To reduce the manpower consumption on box-level annotations, many weakly supervised object detection methods which only require image-level annotations, have been proposed recently. The training process in these methods is formulated into two steps. They firstly train a neural network under weak sup...
| Main Authors: | , , |
|---|---|
| Other Authors: | |
| Format: | Journal Article |
| Language: | English |
| Published: |
2022
|
| Subjects: | |
| Online Access: | https://hdl.handle.net/10356/161623 |
| _version_ | 1826117054212603904 |
|---|---|
| author | Jin, Ruibing Lin, Guosheng Wen, Changyun |
| author2 | School of Computer Science and Engineering |
| author_facet | School of Computer Science and Engineering Jin, Ruibing Lin, Guosheng Wen, Changyun |
| author_sort | Jin, Ruibing |
| collection | NTU |
| description | To reduce the manpower consumption on box-level annotations, many weakly supervised object detection methods which only require image-level annotations, have been proposed recently. The training process in these methods is formulated into two steps. They firstly train a neural network under weak supervision to generate pseudo ground truths (PGTs). Then, these PGTs are used to train another network under full supervision. Compared with fully supervised methods, the training process in weakly supervised methods becomes more complex and time-consuming. Furthermore, overwhelming negative proposals are involved at the first step. This is neglected by most methods, which makes the training network biased towards to negative proposals and thus degrades the quality of the PGTs, limiting the training network performance at the second step. Online proposal sampling is an intuitive solution to these issues. However, lacking of adequate labeling, a simple online proposal sampling may make the training network stuck into local minima. To solve this problem, we propose an Online Active Proposal Set Generation (OPG) algorithm. Our OPG algorithm consists of two parts: Dynamic Proposal Constraint (DPC) and Proposal Partition (PP). DPC is proposed to dynamically determine different proposal sampling strategies according to the current training state. PP is used to score each proposal, part proposals into different sets and generate an active proposal set for the network optimization. Through experiments, our proposed OPG shows consistent and significant improvement on both datasets PASCAL VOC 2007 and 2012, yielding comparable performance to the state-of-the-art results. |
| first_indexed | 2024-10-01T04:21:29Z |
| format | Journal Article |
| id | ntu-10356/161623 |
| institution | Nanyang Technological University |
| language | English |
| last_indexed | 2024-10-01T04:21:29Z |
| publishDate | 2022 |
| record_format | dspace |
| spelling | ntu-10356/1616232022-09-12T06:32:54Z Online active proposal set generation for weakly supervised object detection Jin, Ruibing Lin, Guosheng Wen, Changyun School of Computer Science and Engineering School of Electrical and Electronic Engineering Engineering::Electrical and electronic engineering Engineering::Computer science and engineering Weakly Supervised Learning Object Detection To reduce the manpower consumption on box-level annotations, many weakly supervised object detection methods which only require image-level annotations, have been proposed recently. The training process in these methods is formulated into two steps. They firstly train a neural network under weak supervision to generate pseudo ground truths (PGTs). Then, these PGTs are used to train another network under full supervision. Compared with fully supervised methods, the training process in weakly supervised methods becomes more complex and time-consuming. Furthermore, overwhelming negative proposals are involved at the first step. This is neglected by most methods, which makes the training network biased towards to negative proposals and thus degrades the quality of the PGTs, limiting the training network performance at the second step. Online proposal sampling is an intuitive solution to these issues. However, lacking of adequate labeling, a simple online proposal sampling may make the training network stuck into local minima. To solve this problem, we propose an Online Active Proposal Set Generation (OPG) algorithm. Our OPG algorithm consists of two parts: Dynamic Proposal Constraint (DPC) and Proposal Partition (PP). DPC is proposed to dynamically determine different proposal sampling strategies according to the current training state. PP is used to score each proposal, part proposals into different sets and generate an active proposal set for the network optimization. Through experiments, our proposed OPG shows consistent and significant improvement on both datasets PASCAL VOC 2007 and 2012, yielding comparable performance to the state-of-the-art results. Ministry of Education (MOE) Nanyang Technological University Submitted/Accepted version This work is partly supported by an NTU, Singapore Start-up Grant (04INS000338C130) and MOE, Singapore Tier-1 research grants: RG28/18 (S) and RG22/19 (S). 2022-09-12T06:32:53Z 2022-09-12T06:32:53Z 2022 Journal Article Jin, R., Lin, G. & Wen, C. (2022). Online active proposal set generation for weakly supervised object detection. Knowledge-Based Systems, 237, 107726-. https://dx.doi.org/10.1016/j.knosys.2021.107726 0950-7051 https://hdl.handle.net/10356/161623 10.1016/j.knosys.2021.107726 2-s2.0-85120445171 237 107726 en 04INS000338C130 RG28/18 (S) RG22/19 (S) Knowledge-Based Systems © 2021 Elsevier B.V. All rights reserved. This paper was published in Knowledge-Based Systems and is made available with permission of Elsevier B.V. application/pdf |
| spellingShingle | Engineering::Electrical and electronic engineering Engineering::Computer science and engineering Weakly Supervised Learning Object Detection Jin, Ruibing Lin, Guosheng Wen, Changyun Online active proposal set generation for weakly supervised object detection |
| title | Online active proposal set generation for weakly supervised object detection |
| title_full | Online active proposal set generation for weakly supervised object detection |
| title_fullStr | Online active proposal set generation for weakly supervised object detection |
| title_full_unstemmed | Online active proposal set generation for weakly supervised object detection |
| title_short | Online active proposal set generation for weakly supervised object detection |
| title_sort | online active proposal set generation for weakly supervised object detection |
| topic | Engineering::Electrical and electronic engineering Engineering::Computer science and engineering Weakly Supervised Learning Object Detection |
| url | https://hdl.handle.net/10356/161623 |
| work_keys_str_mv | AT jinruibing onlineactiveproposalsetgenerationforweaklysupervisedobjectdetection AT linguosheng onlineactiveproposalsetgenerationforweaklysupervisedobjectdetection AT wenchangyun onlineactiveproposalsetgenerationforweaklysupervisedobjectdetection |