Table 3 Image classification accuracy (%) over multiple models with various input resolutions on our proposed dataset.
From: TCMP-300: A Comprehensive Traditional Chinese Medicinal Plant Dataset for Plant Recognition
Model | Input Resolution | |||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
224 × 224 | 336 × 336 | 448 × 448 | ||||||||||
Acc | F1 | B-Acc | B-F1 | Acc | F1 | B-Acc | B-F1 | Acc | F1 | B-Acc | B-F1 | |
MobileNetV3-Large36 | 81.76 | 80.19 | 80.02 | 81.69 | 83.24 | 81.78 | 81.66 | 83.16 | 84.80 | 83.46 | 83.35 | 84.77 |
ShuffleNetV237 | 82.13 | 80.59 | 80.50 | 82.03 | 84.07 | 82.60 | 82.46 | 83.96 | 85.40 | 84.00 | 83.86 | 85.27 |
EfficientNet-B038 | 84.62 | 83.16 | 83.10 | 84.53 | 85.89 | 84.55 | 84.44 | 85.79 | 86.97 | 85.71 | 85.65 | 86.89 |
ResNet-5033 | 86.51 | 85.23 | 85.13 | 86.45 | 87.59 | 86.31 | 86.22 | 87.51 | 88.92 | 87.81 | 87.70 | 88.88 |
DenseNet-16134 | 87.76 | 86.58 | 86.43 | 87.68 | 87.88 | 86.74 | 86.58 | 87.81 | 87.95 | 86.93 | 87.06 | 87.82 |
ConNeXt-Tiny39 | 88.18 | 87.03 | 86.96 | 88.09 | 87.17 | 85.92 | 85.81 | 87.08 | 86.69 | 85.38 | 85.31 | 86.59 |
RegNet-X35 | 87.75 | 86.46 | 86.40 | 87.66 | 88.78 | 87.66 | 87.59 | 88.72 | 89.54 | 88.84 | 88.74 | 89.87 |
