Table 4 The results of different models on the UCMerced-LandUse dataset.

From: Local feature acquisition and global context understanding network for very high-resolution land cover classification

Method	Pre	Rec	Acc	F1
AlexNet¹⁴^†	0.5729	0.5835	0.5681	0.5529
VGG11⁴⁶^†	0.7477	0.7225	0.7277	0.7188
VGG13⁴⁶^†	0.8406	0.8365	0.8451	0.8344
VGG16⁴⁶^†	0.8499	0.8441	0.8498	0.8365
VGG19⁴⁶^†	0.7596	0.7598	0.7653	0.7448
GoogleNet⁴⁷^†	0.8622	0.8593	0.8638	0.8521
ResNet34⁴⁸^†	0.9197	0.8835	0.8826	0.8782
ResNet50⁴⁸^†	0.9147	0.8772	0.8873	0.8765
ResNet101⁴⁸^†	0.8976	0.8349	0.8498	0.8283
Mobilenetv2⁵⁰^†	0.9503	0.9483	0.9484	0.9475
Mobilenetv3-l⁵⁰^†	0.9198	0.9161	0.9202	0.9147
Mobilenetv3-s⁵⁰^†	0.9424	0.9320	0.9343	0.9314
Shufflenet- × 0.5⁵¹^†	0.9240	0.9241	0.9249	0.9194
Shufflenet- × 1⁵¹^†	0.9758	0.9764	0.9765	0.9740
Shufflenet- × 1.5⁵¹^†	0.9689	0.9654	0.9671	0.9655
Shufflenetv2- × 2⁵¹^†	0.9788	0.9692	0.9718	0.9724
densenet121⁵²^†	0.9897	0.9848	0.9859	0.9865
densenet161⁵²^†	0.9877	0.9828	0.9859	0.9843
densenet169⁵²^†	0.9794	0.9757	0.9765	0.9763
densenet201⁵²^†	0.9804	0.9761	0.9765	0.9777
Efficient-b0-1 k⁵³^†	0.9601	0.9573	0.9577	0.9572
Efficient-b1-1 k⁵³^†	0.9884	0.9884	0.9859	0.9879
Efficient-b2-1 k⁵³^†	0.9893	0.9881	0.9859	0.9879
Efficient-b3-1 k⁵³^†	0.9873	0.9841	0.9859	0.9847
Efficient-b4-1 k⁵³^†	0.9781	0.9765	0.9765	0.9762
Efficient-b5-1 k⁵³^†	0.9728	0.9700	0.9718	0.9697
Efficient-b6-1 k⁵³^†	0.9784	0.9750	0.9765	0.9760
Efficient-b7-1 k⁵³^†	0.9078	0.8725	0.8685	0.8668
efficientV2_l⁵³^†	0.9422	0.9321	0.9343	0.9344
efficientV2_m⁵³^†	0.9201	0.9115	0.9155	0.9099
efficientV2_s⁵³^†	0.9397	0.9272	0.9296	0.9298
Vgg-Vote⁵⁵^‡	0.9485	0.9642	0.9512	0.9567
Convnext-s-1k⁵⁴^†	0.9603	0.9615	0.9624	0.9587
Convnext-b-22k⁵⁴^†	0.9665	0.9701	0.9671	0.9656
Convnext-b-1k⁵⁴^†	0.9740	0.9728	0.9718	0.9711
Convnext-t-1k⁵⁴^†	0.9687	0.9669	0.9671	0.9660
Convnext-l-1k⁵⁴^†	0.9814	0.9816	0.9812	0.9805
Convnext-l-22k⁵⁴^†	0.9699	0.9630	0.9624	0.9631
Convnext-xl-22k⁵⁴^†	0.9695	0.9705	0.9671	0.9678
DFAGCN⁵⁶^‡	–	–	0.9848	–
SNN-VGG-15⁴²^‡	–	–	0.9914	–
MLLD⁵⁷^‡	–	–	0.7776	–
HSL-MINet⁵⁸^‡	–	–	0.8189	–
ViT-b-p16²⁴^†	0.9592	0.9568	0.9577	0.9555
ViT-b-p32²⁴^†	0.9697	0.9606	0.9624	0.9616
ViT-l-p16²⁴^†	0.9603	0.9533	0.9531	0.9550
T2T-VIT-12⁵⁹‡	–	–	0.9910	–
TransResUNet⁶⁰^†	0.9502	0.9492	0.9484	0.9460
BPECN⁶¹^‡	–	–	0.9772	–
SKAL-AlexNet⁶²^‡	–	–	0.9738	–
SKAL-ResNet18⁶²^‡	–	–	0.9952	–
SKAL-GoogleNet⁶²^‡	–	–	0.9940	–
GCSANet⁶³^‡	–	–	0.9832	–
EMTCAL³²^‡	–	–	0.9929	–
SF-MSFormer-ResNet18⁶⁴^‡	–	–	0.9935	–
LGLFormer⁶⁵^‡	–	–	0.9948	–
LFAGCU (ours)	0.9966	0.9960	0.9953	0.9962

Bold indicates the optimal solution, ^‡represents the data results from the reference literature, and ^†represents the experimental results based on the settings of this paper's parameters.

Back to article page

Table 4 The results of different models on the UCMerced-LandUse dataset.

Search

Quick links