Few-shot object detection via baby learning
Few-shot learning is proposed to overcome the problem of scarce training data in novel classes. Recently, few-shot learning has been well adopted in various computer vision tasks such as object recognition and object detection. However, the state-of-the-art (SOTA) methods have less attention to effe...
Saved in:
Published in | Image and vision computing Vol. 120; p. 104398 |
---|---|
Main Authors | , , , , , , |
Format | Journal Article |
Language | English |
Published |
Elsevier B.V
01.04.2022
|
Subjects | |
Online Access | Get full text |
Cover
Loading…
Summary: | Few-shot learning is proposed to overcome the problem of scarce training data in novel classes. Recently, few-shot learning has been well adopted in various computer vision tasks such as object recognition and object detection. However, the state-of-the-art (SOTA) methods have less attention to effectively reuse the information from previous stages. In this paper, we propose a new framework of few-shot learning for object detection. In particular, we adopt Baby Learning mechanism along with the multiple receptive fields to effectively utilize the former knowledge in novel domain. The propoed framework imitates the learning process of a baby through visual cues. The extensive experiments demonstrate the superiority of the proposed method over the SOTA methods on the benchmarks (improve average 7.0% on PASCAL VOC and 1.6% on MS COCO).
•Introduce Baby Learning mechanism into few-shot object detection.•Use multi-receptive fields to capture the novel variance object appearance in FSOD.•Propose FORD+BL method to achieve superior results over the baseline.•Flexibly apply Baby Learning mechanism to other FSOD methods. |
---|---|
ISSN: | 0262-8856 1872-8138 |
DOI: | 10.1016/j.imavis.2022.104398 |