More compact MobileNet

The new 1x1 convolution module for greatly reducing the number of parameters of the compact model such as MobileNet.

General

the initial design architecture:
　　　　　　　　　　　　　　　　　　　　　　　　　　　　　　　　　　　　　　　　　　　　　　　　　　　　　　　　　　　
the modified design architecture:
　　　　　　　　　　　　　　　　　　　　　　　　　　　　　　　　　　　

Required software

pytorch - 0.4
python - 3.5

Hardware required

The code was tested on P100.

Name		Name	Last commit message	Last commit date
Latest commit History 2 Commits
assets		assets
saved_models		saved_models
Knowldege_Distillation_v1.ipynb		Knowldege_Distillation_v1.ipynb
LICENSE		LICENSE
M_5_256.ipynb		M_5_256.ipynb
M_accelerate_6lastlayer-16fold_.ipynb		M_accelerate_6lastlayer-16fold_.ipynb
M_accelerate_6lastlayer-8fold_91.00-softmax.ipynb		M_accelerate_6lastlayer-8fold_91.00-softmax.ipynb
M_accelerate_6lastlayer-8fold_91.41-Copy1.ipynb		M_accelerate_6lastlayer-8fold_91.41-Copy1.ipynb
M_accelerate_6lastlayer-8fold_91.41-Copy2.ipynb		M_accelerate_6lastlayer-8fold_91.41-Copy2.ipynb
M_accelerate_6lastlayer-8fold_91.41-Copy3.ipynb		M_accelerate_6lastlayer-8fold_91.41-Copy3.ipynb
M_accelerate_6lastlayer-8fold_91.41-Copy4.ipynb		M_accelerate_6lastlayer-8fold_91.41-Copy4.ipynb
M_accelerate_6lastlayer-8fold_91.41-Copy5.ipynb		M_accelerate_6lastlayer-8fold_91.41-Copy5.ipynb
M_accelerate_6lastlayer-8fold_91.41.ipynb		M_accelerate_6lastlayer-8fold_91.41.ipynb
M_accelerate_6lastlayer-8fold_91.43-Copy6.ipynb		M_accelerate_6lastlayer-8fold_91.43-Copy6.ipynb
M_accelerate_6lastlayer-8fold_91.67-Copy1.ipynb		M_accelerate_6lastlayer-8fold_91.67-Copy1.ipynb
M_accelerate_6lastlayer-8fold_91.67-Copy2.ipynb		M_accelerate_6lastlayer-8fold_91.67-Copy2.ipynb
M_accelerate_6lastlayer-8fold_91.67-softmax-Copy1.ipynb		M_accelerate_6lastlayer-8fold_91.67-softmax-Copy1.ipynb
M_accelerate_6lastlayer-8fold_91.67-softmax-Copy10.ipynb		M_accelerate_6lastlayer-8fold_91.67-softmax-Copy10.ipynb
M_accelerate_6lastlayer-8fold_91.67-softmax-Copy11.ipynb		M_accelerate_6lastlayer-8fold_91.67-softmax-Copy11.ipynb
M_accelerate_6lastlayer-8fold_91.67-softmax-Copy12.ipynb		M_accelerate_6lastlayer-8fold_91.67-softmax-Copy12.ipynb
M_accelerate_6lastlayer-8fold_91.67-softmax-Copy2.ipynb		M_accelerate_6lastlayer-8fold_91.67-softmax-Copy2.ipynb
M_accelerate_6lastlayer-8fold_91.67-softmax-Copy3.ipynb		M_accelerate_6lastlayer-8fold_91.67-softmax-Copy3.ipynb
M_accelerate_6lastlayer-8fold_91.67-softmax-Copy4.ipynb		M_accelerate_6lastlayer-8fold_91.67-softmax-Copy4.ipynb
M_accelerate_6lastlayer-8fold_91.67-softmax-Copy5.ipynb		M_accelerate_6lastlayer-8fold_91.67-softmax-Copy5.ipynb
M_accelerate_6lastlayer-8fold_91.67-softmax-Copy6.ipynb		M_accelerate_6lastlayer-8fold_91.67-softmax-Copy6.ipynb
M_accelerate_6lastlayer-8fold_91.67-softmax-Copy7.ipynb		M_accelerate_6lastlayer-8fold_91.67-softmax-Copy7.ipynb
M_accelerate_6lastlayer-8fold_91.67-softmax-Copy8.ipynb		M_accelerate_6lastlayer-8fold_91.67-softmax-Copy8.ipynb
M_accelerate_6lastlayer-8fold_91.67-softmax-Copy9.ipynb		M_accelerate_6lastlayer-8fold_91.67-softmax-Copy9.ipynb
M_accelerate_6lastlayer-8fold_91.67-softmax.ipynb		M_accelerate_6lastlayer-8fold_91.67-softmax.ipynb
M_accelerate_6lastlayer-8fold_91.67.ipynb		M_accelerate_6lastlayer-8fold_91.67.ipynb
M_accelerate_6lastlayer.ipynb		M_accelerate_6lastlayer.ipynb
M_accelerate_7lastlayer-8fold_90.73-Copy1.ipynb		M_accelerate_7lastlayer-8fold_90.73-Copy1.ipynb
M_accelerate_ALL-Copy1.ipynb		M_accelerate_ALL-Copy1.ipynb
M_accelerate_ALL.ipynb		M_accelerate_ALL.ipynb
M_accelerate_K_65.ipynb		M_accelerate_K_65.ipynb
M_accelerate_K_9.ipynb		M_accelerate_K_9.ipynb
Main_02.ipynb		Main_02.ipynb
Main_1.ipynb		Main_1.ipynb
Main_scripts_00.ipynb		Main_scripts_00.ipynb
Main_scripts_01.ipynb		Main_scripts_01.ipynb
Main_scripts_03.ipynb		Main_scripts_03.ipynb
Main_scripts_04.ipynb		Main_scripts_04.ipynb
Main_scripts_04_bs512.ipynb		Main_scripts_04_bs512.ipynb
MobileNet_V2_Raw_1.ipynb		MobileNet_V2_Raw_1.ipynb
Modify_5-Copy1.ipynb		Modify_5-Copy1.ipynb
Modify_5.ipynb		Modify_5.ipynb
NEW_ALL#91.50#FiexedSeedVector#250epoch.ipynb		NEW_ALL#91.50#FiexedSeedVector#250epoch.ipynb
NEW_ALL#92.40#后6层全代替#250epoch#2.53MB.ipynb		NEW_ALL#92.40#后6层全代替#250epoch#2.53MB.ipynb
NEW_ALL#92.51#NoAutoScale#250epoch.ipynb		NEW_ALL#92.51#NoAutoScale#250epoch.ipynb
NEW_ALL#92.91#250epoch.ipynb		NEW_ALL#92.91#250epoch.ipynb
README.md		README.md
Raw_2.ipynb		Raw_2.ipynb
Raw_3_Dense_2_1DConv.ipynb		Raw_3_Dense_2_1DConv.ipynb
Raw_MobileNet_V1-Copy1.ipynb		Raw_MobileNet_V1-Copy1.ipynb
Raw_MobileNet_V1-Copy1_50%.ipynb		Raw_MobileNet_V1-Copy1_50%.ipynb
Raw_MobileNet_V1-Copy1_75%-80epoch_91.81.ipynb		Raw_MobileNet_V1-Copy1_75%-80epoch_91.81.ipynb
Raw_MobileNet_V1-Copy1_75%-Copy1.ipynb		Raw_MobileNet_V1-Copy1_75%-Copy1.ipynb
Raw_MobileNet_V1-Copy1_75%.ipynb		Raw_MobileNet_V1-Copy1_75%.ipynb
Raw_MobileNet_V1-Copy2.ipynb		Raw_MobileNet_V1-Copy2.ipynb
Raw_MobileNet_V1-Copy3.ipynb		Raw_MobileNet_V1-Copy3.ipynb
Raw_MobileNet_V1-Copy4.ipynb		Raw_MobileNet_V1-Copy4.ipynb
Raw_MobileNet_V1-Copy4_Acc_92.99%.ipynb		Raw_MobileNet_V1-Copy4_Acc_92.99%.ipynb
Raw_MobileNet_V1-Copy5.ipynb		Raw_MobileNet_V1-Copy5.ipynb
Raw_MobileNet_V1-Copy6.ipynb		Raw_MobileNet_V1-Copy6.ipynb
Raw_MobileNet_V1-Copy7.ipynb		Raw_MobileNet_V1-Copy7.ipynb
Raw_MobileNet_V1.ipynb		Raw_MobileNet_V1.ipynb
Train_Info.txt		Train_Info.txt
mobilenets.py		mobilenets.py

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Repository files navigation

More compact MobileNet

General

Required software

Hardware required

About

Uh oh!

Releases

Packages

Uh oh!

Contributors

Uh oh!

Languages

Folders and files

Latest commit

History

Repository files navigation

More compact MobileNet

General

Required software

Hardware required

About

Resources

License

Uh oh!

Stars

Watchers

Forks

Releases

Packages 0

Uh oh!

Contributors

Uh oh!

Languages

Packages