Conference item
Many-shot from low-shot: learning to annotate using mixed supervision for object detection
- Abstract:
- Object detection has witnessed significant progress by relying on large, manually annotated datasets. Annotating such datasets is highly time consuming and expensive, which motivates the development of weakly supervised and few-shot object detection methods. However, these methods largely underperform with respect to their strongly supervised counterpart, as weak training signals often result in partial or oversized detections. Towards solving this problem we introduce, for the first time, an online annotation module (OAM) that learns to generate a many-shot set of reliable annotations from a larger volume of weakly labelled images. Our OAM can be jointly trained with any fully supervised two-stage object detection method, providing additional training annotations on the fly. This results in a fully end-to-end strategy that only requires a low-shot set of fully annotated images. The integration of the OAM with Fast(er) R-CNN improves their performance by 17 % mAP, 9 % AP50 on PASCAL VOC 2007 and MS-COCO benchmarks, and significantly outperforms competing methods using mixed supervision.
- Publication status:
- Published
- Peer review status:
- Reviewed (other)
Actions
Access Document
- Files:
-
-
(Preview, Accepted manuscript, 30.1MB, Terms of use)
-
- Publisher copy:
- 10.1007/978-3-030-58598-3_3
Authors
- Publisher:
- Springer Nature
- Host title:
- Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
- Journal:
- Proceedings of the 16th European Conference on Computer Vision (ECCV 2020) More from this journal
- Volume:
- 12353
- Pages:
- 35-50
- Series:
- Lecture Notes in Computer Science
- Publication date:
- 2020-11-07
- Acceptance date:
- 2020-07-02
- Event title:
- European Conference on Computer Vision (ECCV 2020)
- Event location:
- Online
- Event website:
- https://eccv2020.eu/
- Event start date:
- 2021-08-23
- Event end date:
- 2021-08-28
- DOI:
- EISSN:
-
1611-3349
- ISSN:
-
0302-9743
- EISBN:
- 978-3-030-58598-3
- ISBN:
- 9783030585976
- Language:
-
English
- Keywords:
- Pubs id:
-
1151096
- Local pid:
-
pubs:1151096
- Deposit date:
-
2021-01-05
Terms of use
- Copyright holder:
- Springer
- Copyright date:
- 2020
- Rights statement:
- © Springer Nature Switzerland AG 2020
- Notes:
- This paper was presented at the 16th European Conference on Computer Vision (ECCV 2020), 23rd - 28th August 2020. This is the accepted manuscript version of the article. The final version is available from Springer at: https://doi.org/10.1007/978-3-030-58598-3_3
If you are the owner of this record, you can report an update to it here: Report update to this record