notesum.ai

Published at November 2

Covariance-based Space Regularization for Few-shot Class Incremental Learning

cs.CV

cs.AI

Released Date: November 2, 2024

Authors: Yijie Hu¹, Guanyu Yang¹, Zhaorui Tan¹, Xiaowei Huang², Kaizhu Huang³, Qiu-Feng Wang¹

Aff.: ¹Xi'an Jiaotong-Liverpool University; ²Liverpool University; ³Duke Kunshan University

Arxiv: http://arxiv.org/abs/2411.01172v1

Refer to caption

Methods		Accuracy in each session (%) $\uparrow$									Avg Acc.	Final Improv.
Methods		0	1	2	3	4	5	6	7	8	Avg Acc.	Final Improv.
	iCaRL [26]	61.31	46.32	42.94	37.63	30.49	24.00	20.89	18.80	17.21	33.29	+39.20
	NCM [11]	61.31	47.80	39.30	31.90	25.70	21.40	18.70	17.20	14.17	30.83	+42.24
	Data-free Replay [22]	71.84	67.12	63.21	59.77	57.01	53.95	51.55	49.52	48.21	58.02	+8.20
	CEC [18]	72.00	66.83	62.97	59.43	56.70	53.73	51.19	49.24	47.63	57.75	+8.78
	MetaFSCIL [5]	72.04	67.94	63.77	60.29	57.58	55.16	52.90	50.79	49.19	58.85	+7.22
	C-FSCIL [10]	76.40	71.14	66.46	63.29	60.42	57.46	54.78	53.11	51.41	61.61	+5.00
	LIMIT [51]	73.81	72.09	67.87	63.89	60.70	57.77	55.67	53.52	51.23	61.84	+5.18
	CE	75.65	70.45	66.09	62.16	58.96	55.92	53.08	51.05	49.39	60.31	+7.02
	CE-Ours	75.77	70.71	66.53	63.33	60.42	57.46	54.61	52.48	50.65	61.33	+5.76
	FACT^∗ [49]	76.25	70.91	66.41	62.79	59.45	56.22	53.37	51.21	49.48	60.68	+6.93
	FACT^∗-Ours	75.65	70.63	66.79	63.20	59.72	56.68	53.78	51.53	50.01	60.89	+6.40
	SAVC^∗ [29]	80.68	75.89	71.54	67.80	64.85	61.42	58.38	56.43	54.91	65.77	+1.50
	SAVC^∗-Ours	80.90	75.89	71.80	68.59	65.86	62.41	59.33	57.71	56.41	66.54