Skip to main content

Table 6 Performance comparison of the DilatedSkinNet with UNet, SegNet, and DeepLabv3+ on ISIC 2016–2018 test sets

From: Automatic lesion segmentation using atrous convolutional deep neural networks in dermoscopic skin cancer images

Methods

ISIC 2016

ISIC 2017

ISIC 2018

 

ACC

JAC

DICE

Time (s)

ACC

JAC

DICE

Time (s)

ACC

JAC

DICE

Time (s)

U-Net [22]

0.854

0.798

0.832

11

0.764

0.687

0.696

10

0.842

0.793

0.815

24

SegNet [21]

0.908

0.813

0.907

16

0.822

0.679

0.818

16

0.880

0.730

0.879

33

DeepLabv3+ [23]

0.952

0.892

0.952

16

0.878

0.730

0.881

19

0.939

0.888

0.941

20

DilatedSkinNet

0.950

0.904

0.949

10

0.888

0.818

0.884

9

0.942

0.891

0.942

14

  1. Time is in seconds on test sets
  2. The higher values are marked in bold
Back to article page

BMC Medical Imaging

ISSN: 1471-2342

Contact us