notesum.ai

Published at December 3

cs.CV

cs.MM

Released Date: December 3, 2024

Authors: Ze Zhang¹, Enyuan Zhao¹, Ziyi Wan¹, Jie Nie¹, Xinyue Liang¹, Lei Huang¹

Aff.: ¹Ocean University of China

*General Methods
Method	Training	Prompt	CE	RMSE	OA(%)	AA(%)
SAN	$\checkmark$	×	1.1847	-	49.35	49.28
MCAN	$\checkmark$	×	1.0471	-	54.13	54.21
DVQA	$\checkmark$	×	1.0081	-	55.82	55.91
*Remote Sensing Methods
RSVQA	$\checkmark$	×	1.1305	-	49.16	49.07
RSIVQA	$\checkmark$	Seg	0.8344	0.1406	59.48	59.89
EarthVQA	$\checkmark$	Seg	0.8121	0.1387	61.72	62.35
*Multimodal Large Language Models
LLava	×	×	-	-	7.46	7.98
LLava	×	Text	-	-	21.33	20.45
GeoChat	×	×	-	-	12.71	12.38
GeoChat	×	Text	-	-	31.49	29.56
TAF(ours)	$\checkmark$	Tamper	0.7677	0.0889	65.04	65.01