Table 2 Comparison of recognition results on KTH dataset with different approaches. Note that the recognition rate reported by Jhuang et al. [20] is obtained on video sequences from scenarios 1 and 4 only. Other reported rates are on all scenarios.
Method | Classification | Recognition rate % |
|---|---|---|
MMD | Unsupervised | 94.4 |
SVM by charac. Kernel | Supervised | 93.1 |
Schindler and Van Gool [24] | Supervised | 92.7 |
Jhuang et al. [20] | Supervised | 91.7 |
Nowozin et al. [31] | Supervised | 87 |
Wong and Cipolla [32] | Supervised | 86.6 |
Danafar and Gheissari [19] | Supervised | 85.3 |
Niebles et al. [11] | Unsupervised | 83.3 |
Dollár et al. [21] | Supervised | 81.2 |
Schüldt et al. [2] | Supervised | 71.7 |