GitHub - 729593736/CoAtNet: This is a classification task based on CIFAR10，Accuracy is about 87%（without pre-training），The net is CoAtNet（0-5，total coatnet family），with the confusion matrix and acc&loss visualization

729593736 / CoAtNet Public

Notifications You must be signed in to change notification settings
Fork 2
Star 10

This is a classification task based on CIFAR10，Accuracy is about 87%（without pre-training），The net is CoAtNet（0-5，total coatnet family），with the confusion matrix and acc&loss visualization

10 stars 2 forks Branches Tags Activity

Star

Notifications

Branches Tags

Name		Name	Last commit message	Last commit date
Latest commit History 4 Commits
__pycache__		__pycache__
cifar-10-batches-py		cifar-10-batches-py
结果images		结果images
Cifar10有验证集.ipynb		Cifar10有验证集.ipynb
Cifar10直接练.ipynb		Cifar10直接练.ipynb
Readme.txt		Readme.txt
cifar10.h5		cifar10.h5
coatnet.py		coatnet.py
dataset.py		dataset.py
debug.log		debug.log
draw_convnet.py		draw_convnet.py

Repository files navigation

注意：
如果”cifar10直接练“找不到cifar10就把CIFAR10函数里面的路径改一下，
改到你有cifar10的地方，例如root="~/data",然后把cifar10仍data里，或者简单的就是
加个download=True的参数

如果"cifar10有验证集"找不到数据集就在dataset.py里面去找路径，对应上就行了


CoAtNet是谷歌在21年一月份提出的一个模型，相较于ViT有很好的泛化性



2021.12.15-zxy