EZCrop: Energy-Zoned Channels for Robust Output Pruning

作者：Rui Lin, Jie Ran, Dongpeng Wang, King Hung Chiu, Ngai Wong · 发表于：2022 IEEE/CVF Winter Conference on Applications of Computer Vision (WACV) · 年份：2022 · DOI：10.1109/wacv51458.2022.00365 · 被引用次数：6 · 研究领域：Advanced Neural Network Applications、Animal Vocal Communication and Behavior、Hand Gesture Recognition Systems

Recent results have revealed an interesting observation in a trained convolutional neural network (CNN), namely, the rank of a feature map channel matrix remains surprisingly constant despite the input images. This has led to an effective rank-based channel pruning algorithm [23], yet the constant rank phenomenon remains mysterious and unexplained. This work aims at demystifying and interpreting such rank behavior from a frequency-domain perspective, which as a bonus suggests an extremely efficient Fast Fourier Transform (FFT)-based metric for measuring channel importance without explicitly computing its rank. We achieve remarkable CNN channel pruning based on this analytically sound and computationally efficient metric, and adopt it for repetitive pruning to demonstrate robustness via our scheme named Energy-Zoned Channels for Robust Output Pruning (EZCrop), which shows consistently better results than other state-of-the-art channel pruning methods. The codes and Appendix are publicly available at: https://github.com/ruilin0212/EZCrop.