Engagement Classification on DAiSEE
69.06AccuracyPriorNet
Evaluation results
| Method | Links | |
|---|---|---|
| PriorNetPrinciple=Prior-guided2026.05 | 69.06 | |
| EfficientNet + LSTMPrinciple=Hybrid temporal2026.05 | 67.48 | |
| Affective/behavioral features + Ordinal TCNPrinciple=Ordinal features2026.05 | 67.4 | |
| Supervised Contrastive Ordinal TCNPrinciple=Contrastive ordinal features2026.05 | 67.32 | |
| EfficientNet + Bi-LSTMPrinciple=Hybrid temporal2026.05 | 66.39 | |
| Self-Supervised FMAEPrinciple=Self-supervised2026.05 | 64.74 | |
| EfficientNet + TCNPrinciple=Hybrid temporal2026.05 | 64.67 | |
| PANet + STformerPrinciple=Attention-based2026.05 | 64 | |
| ResNet + TCNPrinciple=Hybrid temporal2026.05 | 63.9 | |
| ShuffleNetPrinciple=End-to-end2026.05 | 63.9 | |
| 3D DenseNet with AttentionPrinciple=Attention-based2026.05 | 63.59 | |
| Behavioral features + Neural Turing MachinePrinciple=Behavioral features2026.05 | 61.3 | |
| RefEIP / ModernTCNPrinciple=Temporal features2026.05 | 61.2 | |
| 2D ResNet + LSTMPrinciple=Hybrid temporal2026.05 | 61.15 | |
| Behavioral features + LSTM with attentionPrinciple=Sequential features2026.05 | 60 | |
| 3D-CNN + TCNPrinciple=Hybrid temporal2026.05 | 59.97 | |
| ResNet-50 with LSTM with AttentionPrinciple=Attention-based2026.05 | 58.84 | |
| Hybrid R(2+1)D and spatio-temporal blockPrinciple=Hybrid temporal2026.05 | 58.62 | |
| Long-term Recurrent CNNPrinciple=End-to-end2026.05 | 57.9 | |
| 3D-CNN + LSTMPrinciple=Hybrid temporal2026.05 | 56.6 | |
| HRV + Random ForestPrinciple=Remote physiological features2026.05 | 54.49 | |
| Inflated 3D-CNNPrinciple=End-to-end2026.05 | 52.4 | |
| 3D CNNPrinciple=End-to-end2026.05 | 48.6 |