Hi author, i am interested in how the improved dense trajectory data used in the model. because i can't find the IDT features in the npy file and x = tf.placeholder("float", [None, n_frames ,n_detection, n_input]) in the original model doesn't contain these features.
Hi author, i am interested in how the improved dense trajectory data used in the model. because i can't find the IDT features in the npy file and x = tf.placeholder("float", [None, n_frames ,n_detection, n_input]) in the original model doesn't contain these features.