RelationNet

Learning to Compare: Relation Network for Few-Shot Learning

Introduction

Name:	Relation Network
Embed.:	Conv64F
Type:	Metric
Venue:	CVPR'18
Codes:	LearningToCompare_FSL

• We use ResNet-18 with last_stride = 1 for experiments of Table.2.

Cite this work with:

@inproceedings{DBLP:conf/cvpr/SungYZXTH18,
  author    = {Flood Sung and
               Yongxin Yang and
               Li Zhang and
               Tao Xiang and
               Philip H. S. Torr and
               Timothy M. Hospedales},
  title     = {Learning to Compare: Relation Network for Few-Shot Learning},
  booktitle = {2018 {IEEE} Conference on Computer Vision and Pattern Recognition,
               {CVPR} 2018, Salt Lake City, UT, USA, June 18-22, 2018},
  pages     = {1199--1208},
  year      = {2018},
  url       = {http://openaccess.thecvf.com/content_cvpr_2018/html/Sung_Learning_to_Compare_CVPR_2018_paper.html},
  doi       = {10.1109/CVPR.2018.00131}
}

---

Results and Models

Classification

	Embedding	📖 miniImageNet (5,1)	💻 miniImageNet (5,1)	📖miniImageNet (5,5)	💻 miniImageNet (5,5)	📝 Comments
1	Conv64F	-	51.52 ± 0.37 ⬇️ 📋	-	66.49 ± 0.29 ⬇️ 📋	Table.2
2	ResNet12	-	55.22 ± 0.39 ⬇️ 📋	-	69.25 ± 0.31 ⬇️ 📋	Table.2
3	ResNet18	-	53.98 ± 0.37 ⬇️ 📋	-	71.27 ± 0.31 ⬇️ 📋	Table.2

	Embedding	📖 tieredImageNet (5,1)	💻 tieredImageNet (5,1)	📖tieredImageNet (5,5)	💻 tieredImageNet (5,5)	📝 Comments
1	Conv64F	-	54.37 ± 0.44 ⬇️ 📋	-	71.93 ± 0.35 ⬇️ 📋	Table.2