Table 3 Accuracy (in percentage \(\%\)) with various DNN-based classification schemes for TrashBox dataset.
From: Enhancing trash classification in smart cities using federated deep learning
Technique | Accuracy | ||
|---|---|---|---|
Training | Validation | Test | |
ResNeXt-10133 | 97.86 | 97.15 | 89.62 |
ResNeXt-5033 | 97.66 | 95.18 | 88.68 |
ShuffleNetV236 | 88.48 | 89.99 | 82.91 |
ResNet-3432 | 96.31 | 95.25 | 87.07 |
ResNet-5032 | 96.98 | 95.06 | 87.62 |
ResNet-10132 | 97.31 | 95.82 | 87.24 |
MobileNetV237 | 94.55 | 94.11 | 86.24 |
MobileNetV3-Large35 | 92.74 | 93.28 | 85.13 |
MobileNetV3-Small35 | 85.60 | 87.96 | 81.80 |
GoogLeNet38 | 93.41 | 94.23 | 86.18 |
