Skip to main content
. 2022 Mar 11;224:104534. doi: 10.1016/j.chemolab.2022.104534

Table 1.

Architectural variation in different ResNet models [33].

Layer name Output Size 18 layer 34 layer 50 layer 101 layer 152 layer
Convol_1 112 ​× ​112 7 ​× ​7,64,stride ​= ​2
Convol_2 56 ​× ​56 3 ​× ​3, Max pool, stride ​= ​2
[3x3,643x3,64]x ​2 [3x3,643x3,64]x ​3 [1x1,643x3,641x1,256]x ​3 [1x1,643x3,641x1,256]x ​3 [1x1,643x3,641x1,256]x ​3
Convol_3 28 ​× ​28 [3x3,1283x3,128]x ​2 [3x3,1283x3,128]x ​4 [1x1,1283x3,1281x1,512]x ​4 [1x1,1283x3,1281x1,512]x ​4 [1x1,1283x3,1281x1,512]x ​8
Convol_4 14 ​× ​14 [3x3,2563x3,256]x ​2 [3x3,2563x3,256] ​x ​6 [1x1,2563x3,2561x1,1024]x ​6 [1x1,2563x3,2561x1,1024]x ​23 [1x1,2563x3,2561x1,1024]x ​36
Convol_5 7 ​× ​7 [3x3,5123x3,512]x ​2 [3x3,5123x3,512] ​x ​3 [1x1,5123x3,5121x1,2048]x ​3 [1x1,5123x3,5121x1,2048]x ​3 [1x1,5123x3,5121x1,2048]x ​3
1 ​× ​1 Average Pooling 1000, Softmax function