journal menu
![[Figure 3]](tu5018fig3mag.jpg)
|
|
Figure 3
(a) The ResNet block used in this study and (b) the ResNet-18 architecture. In both parts, the orange, purple, green, red, gray, blue and yellow layers represent 1D Convolutional layers, Dropout layers, Layer Normalization layers, Gaussian error linear unit (GELU) activation functions, 1D Max Pooling layers, 1D Adaptive Max Pooling layers and Fully Connected layers, respectively. The white circles with `+' signs in them represent the addition of two layers. In part (b), each ResNet block was repeated twice, shown by the `2×' next to each block. `nc' and `ks' represent the number of channels and the kernel stride for each convolutional layer. |
![[Figure 3]](tu5018fig3.jpg)
 | JOURNAL OF APPLIED CRYSTALLOGRAPHY |
ISSN: 1600-5767
access
