Research Article
Deep Learning with Taxonomic Loss for Plant Identification
Table 5
Accuracies of two neural networks trained by taxonomic loss with different taxonomic hierarchy.
| Dataset | Neural network | Taxonomic hierarchy | Accuracy (%) | Family | Genus | Species |
| PlantCLEF 2015 | Inception-ResNet-v2 [11] | F-G-S | 83.36 | 76.85 | 70.38 | F-S | 82.04 | 76.10 | 69.36 | G-S | 81.48 | 75.82 | 68.94 | S | 80.66 | 74.57 | 67.93 |
| PlantCLEF 2017 | ShuffleNet-v2 [25] | F-G-S | 66.13 | 60.73 | 53.12 | F-S | 64.08 | 57.71 | 50.03 | G-S | 64.26 | 59.11 | 51.64 | S | 61.94 | 57.12 | 49.96 |
|
|