Table 4 Summary of performance metrics for different deep learning architectures applied to the validation set of the benchmark.
From: A multi-filter deep transfer learning framework for image-based autism spectrum disorder detection
DL Architecture | Scale | Length | Use DA | ACC | F1 | AUC | EER | ACE | Recall | Precision | Specificity | FP | FN | TP | TN |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
AffectNet | None | 1000 | False | 82.0 | 81.63 | 90.60 | 18.0 | 18.0 | 80.0 | 83.33 | 84.0 | 8 | 10 | 40 | 42 |
AffectNet + FW | None | 167 | False | 87.0 | 86.60 | 90.60 | 16.0 | 13.0 | 84.0 | 89.36 | 90.0 | 5 | 8 | 42 | 45 |
AlexNet | None | 1000 | False | 82.0 | 81.25 | 88.08 | 22.0 | 18.0 | 78.0 | 84.78 | 86.0 | 7 | 11 | 39 | 43 |
AlexNet + FW | Robust | 43 | False | 82.0 | 82.69 | 90.40 | 18.0 | 18.0 | 86.0 | 79.63 | 78.0 | 11 | 7 | 43 | 39 |
ResNet-50 | None | 1000 | False | 72.0 | 70.83 | 83.56 | 24.0 | 28.0 | 68.0 | 73.91 | 76.0 | 12 | 16 | 34 | 38 |
ResNet-50 + FW | Standard | 36 | False | 80.0 | 80.00 | 86.76 | 20.0 | 20.0 | 80.0 | 80.00 | 80.0 | 10 | 10 | 40 | 40 |
VGG16 | None | 1000 | False | 70.0 | 66.67 | 78.44 | 28.0 | 30.0 | 60.0 | 75.00 | 80.0 | 10 | 20 | 30 | 40 |
VGG16 + FW | Robust | 31 | False | 77.0 | 76.77 | 83.84 | 22.0 | 23.0 | 76.0 | 77.55 | 78.0 | 11 | 12 | 38 | 39 |
VGG19 | None | 1000 | False | 68.0 | 63.64 | 76.36 | 30.0 | 32.0 | 56.0 | 73.68 | 80.0 | 10 | 22 | 28 | 40 |
VGG19 + FW | Standard | 38 | False | 74.0 | 73.47 | 79.76 | 28.0 | 26.0 | 72.0 | 75.00 | 76.0 | 12 | 14 | 36 | 38 |
ViT | None | 1000 | False | 81.0 | 80.81 | 92.36 | 18.0 | 19.0 | 80.0 | 81.63 | 82.0 | 9 | 10 | 40 | 41 |
ViT + FW | None | 130 | False | 85.0 | 85.15 | 91.00 | 16.0 | 15.0 | 86.0 | 84.31 | 84.0 | 8 | 7 | 43 | 42 |
ViTFER | None | 1000 | False | 80.0 | 79.17 | 89.68 | 20.0 | 20.0 | 76.0 | 82.61 | 84.0 | 8 | 12 | 38 | 42 |
ViTFER + FW | Robust | 213 | False | 86.0 | 84.78 | 92.64 | 14.0 | 14.0 | 78.0 | 92.86 | 94.0 | 3 | 11 | 39 | 47 |
ViTSwin | None | 1000 | False | 81.0 | 80.81 | 91.68 | 18.0 | 19.0 | 80.0 | 81.63 | 82.0 | 9 | 10 | 40 | 41 |
ViTSwin + FW | None | 186 | False | 87.0 | 86.60 | 91.96 | 16.0 | 13.0 | 84.0 | 89.36 | 90.0 | 5 | 8 | 42 | 45 |
