Table 2 The results of quantitative comparison.

From: Research on recognition of diabetic retinopathy hemorrhage lesions based on fine tuning of segment anything model

Model

OIA-DDR

JYFY-HE

IDRiD

nIou

nDice

nIou

nDice

nIou

nDice

SAM-ada-Res

0.3154

0.4535

0.4374

0.6040

0.4182

0.5658

AgileFormer

0.3009

0.4417

0.4108

0.5785

0.4041

0.5517

G-CASCADE

0.3215

0.4589

0.3603

0.5241

0.4122

0.5622

Rolling-Unet

0.2880

0.4549

0.1929

0.3526

0.2880

0.4367

Mamba-Unet

0.0618

0.1193

0.2966

0.4713

0.2566

0.4126

VM-Unet

0.0295

0.0533

0.2410

0.3728

0.3709

0.5185

Unet

0.3056

0.4246

0.4227

0.5802

0.3976

0.5449

SAM2-Unet

0.2515

0.3820

0.3866

0.5514

0.3953

0.5451

  1. The best result within each column is highlighted by bold, and the second-best is highlighted with italics.