Skip to main content
. 2025 Aug 12;8:1632344. doi: 10.3389/frai.2025.1632344

Table 2.

Classification performance for domain-adversarial neural networks (DANNs) with EfficientNetV2 backbone.

Domain Metric Source-only 1-shot DANN 3-shot DANN 5-shot DANN
BF source Accuracy (%) 94.4 91.1 94.4 93.3
Precision (%) 94.6 91.1 94.5 93.6
Recall (%) 94.4 91.1 94.4 93.3
F1-score (%) 94.4 91.0 94.4 93.3
BF target Accuracy (%) 41.7 56.7 75.0 73.3
Precision (%) 37.6 56.2 76.1 73.4
Recall (%) 41.7 56.7 75.0 73.3
F1-score (%) 38.2 53.4 73.1 73.3
20 × source Accuracy (%) 94.4 90.0 90.0 93.3
Precision (%) 94.6 90.4 90.1 93.5
Recall (%) 94.4 90.0 90.0 93.3
F1-score (%) 94.4 90.0 89.9 93.4
20 × target Accuracy (%) 33.3 54.4 82.2 88.9
Precision (%) 40.3 43.3 83.3 89.9
Recall (%) 33.3 54.4 82.2 88.9
F1-score (%) 28.1 46.8 82.0 88.8
20x-5h source Accuracy (%) 94.4 94.4 94.4 94.4
Precision (%) 94.6 94.6 94.7 94.8
Recall (%) 94.4 94.4 94.4 94.4
F1-score (%) 94.4 94.4 94.5 94.5
20x-5h target Accuracy (%) 40.0 71.7 83.3 83.3
Precision (%) 43.4 80.1 84.1 84.1
Recall (%) 40.0 71.7 83.3 83.3
F1-score (%) 34.1 70.9 81.3 81.3

Results are presented for source-only training and domain-adversarial training with 1-shot, 3-shot, and 5-shot labeled samples. Bold values indicate the highest accuracy for each domain and setting.