Table 2.
Test accuracy across all species computed by the five classification models.
| Categories | InceptionV4 | DenseNet121 | ResNet50 | MobileNetV2 | ShuffleNetV2 |
|---|---|---|---|---|---|
| Bare land | 76.7 | 78.3 | 83.2 | 79.8 | 81.3 |
| Cotton | 62.7 | 76.2 | 84.7 | 69.5 | 81.4 |
| Maize | 91.1 | 93.1 | 92.1 | 92.9 | 92 |
| Peanut | 59.0 | 64.7 | 73.4 | 68.3 | 65.5 |
| Rapeseed | 89.2 | 96.7 | 93.5 | 88.9 | 93.2 |
| Rice | 93.5 | 87.0 | 88.9 | 95.5 | 93 |
| Sorghum | 59.5 | 73.0 | 73.0 | 70.3 | 75.7 |
| Soybean | 85.5 | 83.5 | 85.4 | 79.4 | 82.1 |
| Sunflower | 79.5 | 65.9 | 72.7 | 56.8 | 72.7 |
| Tobacco | 72.9 | 65.7 | 71.4 | 81.4 | 82.9 |
| Vegetable | 81.1 | 80.4 | 81.8 | 79.7 | 81.3 |
| Wheat | 88.4 | 84.5 | 91.7 | 96.0 | 89.3 |
| Average accuracy | 86.2 | 86 | 87.9 | 87.5 | 87.5 |
Numbers in bold represent the best classification accuracy for each cropland type.