Lip Reading on LRW original (test)
88.5Top-1 AccuracyEnsemble
Evaluation Results
| Method | Links | |
|---|---|---|
| EnsembleStrategy=Ensemble of distillation models2020.07 | 88.5 | |
| ResNet-18 + MS-TCN - Student 3Backbone=ResNet-18, Self-distillation iteration=32020.07 | 87.9 | |
| ResNet-18 + MS-TCN - Student 2Backbone=ResNet-18, Self-distillation iteration=22020.07 | 87.8 | |
| ResNet-18 + MS-TCN - Student 4Backbone=ResNet-18, Self-distillation iteration=42020.07 | 87.7 | |
| ResNet-18 + MS-TCN - Student 1Backbone=ResNet-18, Self-distillation iteration=12020.07 | 87.4 | |
| ResNet-18 + MS-TCNBackbone=ResNet-182020.07 | 85.3 | |
| ResNet-18 + MS-TCN - TeacherBackbone=ResNet-18, Mode=Teacher (Baseline)2020.07 | 85.3 | |
| ResNet-18 + BGRU + CutoutBackbone=ResNet-18, Data Augmentation=Cutout2020.07 | 85 | |
| ResNet-18 + BLSTMBackbone=ResNet-182020.07 | 84.3 | |
| 2-stream 3D-CNN + BLSTM2020.07 | 84.1 | |
| ResNet34 + BGRUBackbone=ResNet-342020.07 | 83.4 | |
| ResNet34 + BLSTMBackbone=ResNet-342020.07 | 83 | |
| Seq-to-Seq2020.07 | 76.2 | |
| 3D-CNN2020.07 | 61.1 |