Table 2 Recognition results for the MMI database
From: Multithreading cascade of SURF for facial expression recognition
Method | Accuracy on MMI (%) | Â | |||||
|---|---|---|---|---|---|---|---|
| Â | An | Di | Fe | Ha | Sa | Su | mAP. |
LBP-TOP [12] | 58.1 | 56.3 | 53.6 | 78.6 | 46.9 | 50.0 | 57.2 |
HOE [13] | 46.4 | 58.3 | 33.2 | 62.6 | 60.8 | 65.1 | 55.5 |
HOG 3D [14] | 61.3 | 53.1 | 39.3 | 78.6 | 43.8 | 55.0 | 55.2 |
3D LUT [6] | 43.3 | 55.3 | 56.8 | 71.4 | 28.2 | 77.5 | 47.2 |
ITBN [35] | 46.9 | 54.8 | 57.1 | 71.4 | 65.6 | 62.5 | 59.7 |
LSH-CORF [36] | 59.6 | 71.4 | 62.3 | 68.9 | 70.3 | 75.1 | 61.8 |
3DCNN-DAP [34] | 64.5 | 62.5 | 50.0 | 85.7 | 53.1 | 57.5 | 62.2 |
STM-ExpLet [29] | – | – | – | – | – | – | 65.4 |
McSURF (3T) | 67.2 | 61.7 | 65.1 | 80.4 | 65.4 | 83.9 | 70.6 |
McSURF (OD) | 69.5 | 65.3 | 68.4 | 83.9 | 68.2 | 82.6 | 73.0 |