Skip to main content
. 2021 Dec 11;21(24):8291. doi: 10.3390/s21248291

Table 15.

Comparative analysis of the proposed model with existing methods.

Methods Dataset
Hockey Fight (%) Surveillance Fight (%)
Motion Blobs and Random Forest [14] 82.40 --
VIF [15] 82.90 --
ViF, OViF, AdaBoost, and SVM [16] 87.50 --
Fine-Tune MobileNet [16] 87.00 --
Improved Fisher Vectors [17] 93.70 --
Hough Forests and 2D CNN [19] 94.60 --
HOG3D + KELM [20] 95.05 --
Two streams (Optical Flow + Darknet-19) [32] 98 74
Proposed CNN-LSTM model 98.00 81.05