Table 3 Feature maps of 3D image blocks based classification results by 3D-ResNet, 3D-MobileNetV2, 3D-DenseNet and 3D-SENet
From: Classifying the tracing difficulty of 3D neuron image blocks based on deep learning
Dataset | 3D-ResNet | 3D-MobileNetV2 | 3D-DenseNet | 3D-SENet | ||||
|---|---|---|---|---|---|---|---|---|
Accuracy | F1 | Accuracy | F1 | Accuracy | F1 | Accuracy | F1 | |
Training | 83.50±0.14 | 85.18±0.13 | 79.97±0.18 | 81.89±0.16 | \(\underline{84.69\pm 0.16}\) | \(\underline{86.33\pm 0.14}\) | 82.29±0.05 | 84.16±0.04 |
Test | \(\underline{81.29\pm 0.39}\) | \(\underline{82.78\pm 0.61}\) | 78.40±0.23 | 79.83±0.22 | 72.03±2.47 | 75.50±1.29 | 80.11±0.50 | 81.63±0.92 |