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.
 
 
 
 
 
deng 9a7030a83c 更新 'README.md' 1 year ago
ascend310_infer init 2 years ago
scripts init 2 years ago
src init 2 years ago
README.md 更新 'README.md' 1 year ago
__init__.py init 2 years ago
eval.py init 2 years ago
export.py init 2 years ago
imagenet_to_mindrecord.py init 2 years ago
postprocess.py init 2 years ago
preprocess.py init 2 years ago
requriments.txt init 2 years ago
train.py init 2 years ago

目前计算机视觉卷积神经网络任务中大多都会使用BatchNormalization,用来归一化数据,保持训练的稳定性,这一方法是非常有效的。但是BN也带来了额外的不小的计算开销、模型训练和推理时的行为差异和打破了小批量样本之间的独立性的问题。为此,DeppMind提出了构建Normal Free的模型,用StdConv代替普通的CNN,实现稳定训练和优异的性能。

Python C++ Shell Text other

Contributors (2)