Fig 1.
G-bneck module.
Fig 2.
a. Ghost module module b. Channel attention and spatial attention module c. GhostNet-CBAM network architecture. Where blue indicates the bottleneck layer for stride=1, orange indicates the bottleneck layer for stride=2, and gray indicates the dual attention module. The network as a whole consists of seven stages, CBAM modules, and three fully connected layers. Specific network parameters can be found in Table 1. GhostNet-CBAM network model parameters.
Table 1.
General network structure of GhostNet-CBAM.
Table 2.
Statistics on basic soil information.
Fig 3.
Spectrograms of data from different datasets.
Table 3.
GhostNet-CBAM and GhostNet model prediction results.
Fig 4.
Graph of the results of the MCCV algorithm.
Table 4.
Prediction results for original and optimized data.
Table 5.
PLSR, VGGNet16, and GhostNet-CBAM prediction results.
Fig 5.
Boxplot of model predictions versus reference values.
Fig 6.
Scatter density plot of reference values vs. predictions.