Fig. 6.

The simplified architecture of the proposed Deep Learning model based on hybrid CNN and parallel LSTM to select the deep features of a given frame set (e.g. Golf swing action recognition using spatio-temporal information)