RepSR: Training Efficient VGG-style Super-Resolution Networks with Structural Re-Parameterization and Batch Normalization. (arXiv:2205.05671v1 [cs.CV])

Web: http://arxiv.org/abs/2205.05671

This paper explores training efficient VGG-style super-resolution (SR)
networks with the structural re-parameterization technique. The general
pipeline of re-parameterization is to train networks with multi-branch topology
first, and then merge them into standard 3x3 convolutions for efficient
inference. In this work, we revisit those primary designs and investigate
essential components for re-parameterizing SR networks. First of all, we find
that batch normalization (BN) is important to bring training non-linearity and
improve the final performance. However, BN is typically ignored in …

arxiv cv networks training