Found 1 document(s)
Page 1 of 1
-
Current state-of-the art approaches to action recognition emphasize learning ConvNets on large amounts of training data, using 3D convolutions to process the temporal dimension. This approach is expensive in terms of mem...