Is 3D Convolution with 5D Tensors Really Necessary for Video Analysis?
This addresses efficiency issues for real-time applications like robotics on edge devices, but it is incremental as it builds on existing splitting methods.
The paper tackles the computational expense of 3D convolutions with 5D tensors in video analysis by proposing novel techniques using 2D and 1D convolutions with 4D/3D tensors, resulting in improved speed and accuracy with fewer parameters.
In this paper, we present a comprehensive study and propose several novel techniques for implementing 3D convolutional blocks using 2D and/or 1D convolutions with only 4D and/or 3D tensors. Our motivation is that 3D convolutions with 5D tensors are computationally very expensive and they may not be supported by some of the edge devices used in real-time applications such as robots. The existing approaches mitigate this by splitting the 3D kernels into spatial and temporal domains, but they still use 3D convolutions with 5D tensors in their implementations. We resolve this issue by introducing some appropriate 4D/3D tensor reshaping as well as new combination techniques for spatial and temporal splits. The proposed implementation methods show significant improvement both in terms of efficiency and accuracy. The experimental results confirm that the proposed spatio-temporal processing structure outperforms the original model in terms of speed and accuracy using only 4D tensors with fewer parameters.