notesum.ai

Published at November 10

Layer-Wise Feature Metric of Semantic-Pixel Matching for Few-Shot Learning

cs.CV

cs.AI

Released Date: November 10, 2024

Authors: Hao Tang¹, Junhao Lu¹, Guoheng Huang¹, Ming Li², Xuhang Chen³, Guo Zhong⁴, Zhengguang Tan¹, Zinuo Li⁵

Aff.: ¹School of Computer Science and Technology, Guangdong University of Technology, Guangdong, China; ²Zhejiang Institute of Optoelectronics, Jinhua, China; ³School of Computer Science and Engineering, Huizhou University, Huizhou, China; ⁴School of Information Science and Technology, Guangdong University of Foreign Studies, Guangzhou, China; ⁵School of Physics, Maths and Computing, University of Western Australia, Western Australia, Australia

Arxiv: http://arxiv.org/abs/2411.06363v1

Methods	Backbone	5-way 1-shot	1-way 1-shot
Shot-Free [46]	ResNet12	59.04	77.64
TPN [47]	ResNet12	59.46	75.65
MTL [57]	ResNet12	61.20 $\pm$ 1.80	75.50 $\pm$ 0.80
RFS-simple [18]	ResNet12	62.02 $\pm$ 0.63	79.64 $\pm$ 0.44
AFHN [42]	ResNet18	62.38 $\pm$ 0.72	78.16 $\pm$ 0.56
DC [48]	ResNet18	62.53 $\pm$ 0.19	79.77 $\pm$ 0.19
ProtoNet [49]	ResNet12	62.39 $\pm$ 0.21	80.53 $\pm$ 0.14
MetaOptNet [45]	ResNet12	62.64 $\pm$ 0.82	78.63 $\pm$ 0.46
SimpleShot [19]	ResNet18	62.85 $\pm$ 0.20	80.02 $\pm$ 0.14
MatchNet [50]	ResNet12	63.08 $\pm$ 0.80	75.99 $\pm$ 0.60
CAN [3]	ResNet12	63.85 $\pm$ 0.48	79.44 $\pm$ 0.34
NegMargin [17]	ResNet12	63.85 $\pm$ 0.81	81.57 $\pm$ 0.56
CTM [51]	ResNet18	64.12 $\pm$ 0.82	80.51 $\pm$ 0.13
MLCN [56]	ResNet12	65.54 $\pm$ 0.43	81.63 $\pm$ 0.31
DeepEMD [52]	ResNet12	65.91 $\pm$ 0.82	82.41 $\pm$ 0.56
FEAT [53]	ResNet12	66.78 $\pm$ 0.20	82.05 $\pm$ 0.14
GLFA [44]	ResNet12	67.25 $\pm$ 0.36	82.80 $\pm$ 0.30
MCL-Katz [58]	ResNet12	67.51	83.99
CVET [59]	ResNet12	70.19 $\pm$ 0.46	84.66 $\pm$ 0.29
RENet [4]	ResNet12	67.60 $\pm$ 0.44	82.58 $\pm$ 0.30
MCNet [5]	ResNet18	72.13 $\pm$ 0.43	88.20 $\pm$ 0.23
LWFM-SPM(Ours)	ResNet18	72.22 $\pm$ 0.43	88.22 $\pm$ 0.24