Table 3 Results of ablation study on the public dataset BioHit.

From: An improved lightweight tongue segmentation model with self-attention parallel network and progressive upsampling

Network

Parallel

Eca attention

Progressive upsample

MPA(%)

Segformer

   

95.14

Segformer + Parallele

√

  

97.25

Segformer + Eca Attention

 

√

 

97.06

Segformer + Progressive Upsample

  

√

96.80

Segformer + Parallel + Eca Attention

√

√

 

97.33

PAPU_TonSeg

√

√

√

98.60

Back to article page