Table 1.
Diseases’ characteristics in tea leaf.
Fig 1.
Disease samples of tea leaf.
Fig 2.
The process of DWT.
Fig 3.
Results of 4 level 2D DWT.
Fig 4.
Fusion of frequency and deep features.
Fig 5.
The network structure of WaveLiteNet.
Fig 6.
Structure of Bneck.
Fig 7.
Accuracy curves under different learning rates.
Fig 8.
Comparison curves of accuracy.
Table 2.
Results with model structure optimization.
Table 3.
Performance comparison with different models.
Table 4.
Performance comparison under different loss functions and attention mechanisms.
Fig 9.
AUC-ROC curve for tea disease classification.
Fig 10.
Confusion matrix for MobileNetV3 on the test set.
Fig 11.
Confusion matrix for WaveLiteNet on the test set.
Table 5.
Comparison of WaveLiteNet and MobileNetV3 on low-resource devices.
Fig 12.
Comparison of inference time on different GPUs.