ShuffleNet V2: Practical Guidelines for Efficient CNN Architecture Design
This work addresses the need for more practical and efficient neural network designs for deployment on various platforms, though it is incremental as it builds upon prior work like ShuffleNet.
The paper tackled the problem of designing efficient CNN architectures by moving beyond FLOPs as an indirect metric to evaluate direct metrics like speed, considering factors such as memory access and platform characteristics, resulting in ShuffleNet V2, which achieves state-of-the-art tradeoffs in speed and accuracy as verified by ablation experiments.
Currently, the neural network architecture design is mostly guided by the \emph{indirect} metric of computation complexity, i.e., FLOPs. However, the \emph{direct} metric, e.g., speed, also depends on the other factors such as memory access cost and platform characterics. Thus, this work proposes to evaluate the direct metric on the target platform, beyond only considering FLOPs. Based on a series of controlled experiments, this work derives several practical \emph{guidelines} for efficient network design. Accordingly, a new architecture is presented, called \emph{ShuffleNet V2}. Comprehensive ablation experiments verify that our model is the state-of-the-art in terms of speed and accuracy tradeoff.