You can not select more than 25 topics Topics must start with a chinese character,a letter or number, can include dashes ('-') and can be up to 35 characters long.
 
 
 
 
SFoever 4e5cc966d1 Add readme figures 1 year ago
.github first commit 1 year ago
configs first commit 1 year ago
docs first commit 1 year ago
examples first commit 1 year ago
figures Add readme figures 1 year ago
mindcv first commit 1 year ago
requirements first commit 1 year ago
tests first commit 1 year ago
tutorials first commit 1 year ago
.gitignore first commit 1 year ago
CONTRIBUTING.md first commit 1 year ago
LICENSE.md first commit 1 year ago
README.md 更新 'README.md' 1 year ago
RELEASE.md first commit 1 year ago
config.py first commit 1 year ago
infer.py first commit 1 year ago
package.sh first commit 1 year ago
quick_start.ipynb first commit 1 year ago
quick_start.md first commit 1 year ago
requirements.txt first commit 1 year ago
setup.py first commit 1 year ago
train.py first commit 1 year ago
train_with_func.py first commit 1 year ago
validate.py first commit 1 year ago
validate_with_func.py first commit 1 year ago

从一个相对不受约束的设计空间开始,逐步探索、改进网络设计空间的规则,直到一个由简单的“规则”网络组成的低维设计空间,称之为RegNet。RegNet设计空间的核心很简单:stage的宽度和深度由量化的线性函数决定。与AnyNet相比,RegNet设计空间的模型更简单,更容易解释,好模型的集中度更高。RegNet网络和当时最先进的不同体量网络在ImageNet数据集上进行了比较,体量较大时,RegNet超越了当时的最先进网络EfficientNet的结果。

Jupyter Notebook Python Markdown Text

Contributors (1)