BriefGPT.xyz
Jan, 2021
RepVGG：重塑VGG风格的卷积神经网络
RepVGG: Making VGG-style ConvNets Great Again
HTML
PDF
Xiaohan Ding, Xiangyu Zhang, Ningning Ma, Jungong Han, Guiguang Ding...
TL;DR
本文提出了一种名为RepVGG的卷积神经网络架构，通过结构重参数化技术实现了训练时和推理时架构的解耦，并在ImageNet上取得了超过80％的top-1精度。与现有的EfficientNet和RegNet等模型相比，RepVGG模型具有更快的速度和更好的精度-速度平衡。
Abstract
We present a simple but powerful architecture of
convolutional neural network
, which has a VGG-like inference-time body composed of nothing but a stack of 3x3 convolution and
relu
, while the training-time model h
→