Table 7 Experimental results of different model architectures in landscape element classification and fine-grained PLIR tasks.
From: The analysis of landscape design and plant selection under deep learning
Model architecture | Accuracy of landscape element classification (%) | Accuracy of fine-grained PLIR (%) |
|---|---|---|
FCN-8s | 85.2 | 81.7 |
CNN + FCN (Traditional combination) | 88.1 | 84.3 |
Improved FCN | 89.7 | 86.5 |
FCN + Faster R-CNN | 90.4 | 88.7 |
