Skip to main content
. 2024 Jul 2;14:15170. doi: 10.1038/s41598-024-66234-3

Table 2.

Original ResNet101 and improved ResNet101.

Original ResNet101 Improved ResNet101
Input image (Width=W, Hight=H, Channel=C=3) Input image (Width=W, Hight=H, Channel=3)
Group1: Conv2d (kerenel=7, stirde=2), (W/2, H/2, C=64) Group1: Conv2d (kerenel=7), stirde=2, (W/2, H/2, C=64)
Maxpool (kerenl=3, stride=2), (W/4, H/4, C=64) Maxpool (kerenl=3, stride=2), (W/4, H/4, C=64)
Group2:1×1,643×3,641×1,256×3(W/4,H/4,C=256) Group2:1×1,643×3,641×1,256×3(W/4,H/4,C=256)
Group3:1×1,1283×3,1281×1,512×4(W/8,H/8,C=256) Group3:1×1,1283×3,1281×1,512×4(W/8,H/8,C=256)
Group4:1×1,2563×3,2561×1,1024×23(W/16,H/16,C=512) Group4:BottleNeck_s×23(W/16,H/16,C=1024)
Group5:1×1,5123×3,5121×1,2048×3(W/32,H/32,C=2048) Group5:BottleNeck_d×11×1,5123×3,5121×1,2048×2(W/16,H/16,C=2048)