CAN

Cross Attention Network for Few-shot Classification

Introduction

Name:	CAN
Embed.:	Conv64F
Type:	Metric
Venue:	NeurIPS'17
Codes:	fewshot-CAN

Cite this work with:

@inproceedings{DBLP:conf/nips/HouCMSC19,
  author    = {Ruibing Hou and
               Hong Chang and
               Bingpeng Ma and
               Shiguang Shan and
               Xilin Chen},
  title     = {Cross Attention Network for Few-shot Classification},
  booktitle = {Advances in Neural Information Processing Systems 32: Annual Conference
               on Neural Information Processing Systems 2019, NeurIPS 2019, December
               8-14, 2019, Vancouver, BC, Canada},
  pages     = {4005--4016},
  year      = {2019},
  url       = {https://proceedings.neurips.cc/paper/2019/hash/01894d6f048493d2cacde3c579c315a3-Abstract.html}
}

---

Results and Models

Classification

	Embedding	📖 miniImageNet (5,1)	💻 miniImageNet (5,1)	📖miniImageNet (5,5)	💻 miniImageNet (5,5)	📝 Comments
1	Conv64F	-	55.88 ± 0.38⬇️ 📋	-	70.98 ± 0.30⬇️ 📋	Table.2
2	ResNet12	-	59.82 ± 0.38 ⬇️ 📋	-	76.54 ± 0.29 ⬇️ 📋	Table.2
3	ResNet18	-	60.78 ± 0.40 ⬇️ 📋	-	75.05 ± 0.29 ⬇️ 📋	Table.2 (HW=11)

	Embedding	📖 tieredImageNet (5,1)	💻 tieredImageNet (5,1)	📖tieredImageNet (5,5)	💻 tieredImageNet (5,5)	📝 Comments
1	Conv64F	-	55.96 ± 0.42 ⬇️ 📋	-	70.52 ± 0.35 ⬇️ 📋	Table.2
2	ResNet12	-	70.46 ± 0.43 ⬇️ 📋	-	84.50 ± 0.30 ⬇️ 📋	Table.2
3	ResNet18	-	71.70 ± 0.43 ⬇️ 📋	-	84.61 ± 0.37 ⬇️ 📋	Table.2 (HW=11)