Table 2 Experimental results of the model on the BPPV test set
From: Multimodal deep learning-based diagnostic model for BPPV
Model | Input | Feature fusion | Accuracy |
|---|---|---|---|
BKTDN | Multimodal | Cross-attention module | 81.7% |
BKTDN | Unimodal | Self-attention module | 52.8% |
BKTDN | Unimodal | Concatenate | 39.4% |
BKTDN | Multimodal | Self-attention module | 79.8% |
BKTDN | Multimodal | Concatenate | 73.0% |
TDN | Multimodal | Cross-attention | 79.8% |
TDN | Multimodal | Self-attention | 75.9% |
TDN | Multimodal | Concatenate | 71.1% |
BKTDN | Multimodal | Weighted summation | 73.0% |
BKTDN | Multimodal | Weighted summation+Self-attention | 76.9% |