notesum.ai

Published at November 29

LaVIDE: A Language-Vision Discriminator for Detecting Changes in Satellite Image with Map References

cs.CV

cs.AI

cs.LG

Released Date: November 29, 2024

Authors: Shuguo Jiang¹, Fang Xu¹, Sen Jia², Gui-Song Xia¹

Aff.: ¹School of Computer Science, Wuhan University; ²College of Computer Science and Software Engineering, Shenzhen University

Arxiv: http://arxiv.org/pdf/2411.19758v1

Refer to caption

Methods	Type	DynamicEarthNet		HRSCD		BANDON		SECOND
	Type	F1.	IoU	F1.	IoU	F1.	IoU	F1.	IoU
SETR_PUP [55]	Category	20.8	11.6	5.4	2.8	24.8	14.2	51.2	34.4
SSG2 [16]	Category	17.2	9.4	3.7	1.8	23.7	13.4	47.9	31.5
Segformer [50]	Category	21.2	11.9	5.6	2.8	25.1	14.4	51.2	34.4
SNUNet [17]	Vision	4.4	2.3	49.8	33.2	42.0	26.6	39.6	24.7
CGNet [22]	Vision	16.0	8.7	62.9	45.9	71.3	55.4	65.0	48.1
ChangeFormer [3]	Vision	31.2	18.5	62.1	45.0	69.8	53.6	65.1	48.3
FHD [42]	Vision	32.8	19.6	56.4	39.3	69.0	52.7	64.2	47.3
ChangerEx [18]	Vision	15.3	8.3	20.6	11.5	26.3	15.2	32.1	19.1
Mapformer [4]	Vision	32.0	19.0	62.1	45.0	71.4	55.5	67.3	50.7
LaVIDE	Language-Vision	36.5	22.3	65.2	48.4	72.5	56.9	69.1	52.9