Machine Learning Model for Group Activity Recognition Based on Discriminative Interaction Contextual Relationship

Table 1

Performance of the proposed method compared with the state-of-art methods for group activity recognition accuracy on collective activity dataset [10].

Method/Activities	Crossing (%)	Waiting (%)	Queuing (%)	Walking (%)	Talking (%)	Accuracy (%)
AC [13]	68	69	76	80	99	78.4
RSTV + MRF [23]	76.4	76.7	78.7	36.8	85.7	70.9
AC + FC-CRF [26]	67	84	86	49	75	72.2
Interaction joint context descriptor	68.3	76.2	83.1	80.6	86	78.84
Proposed GAC descriptor	85.2	87.1	87	84.4	98.2	88.38