Skip to content

Source code for "Mean teachers are better role models: Weight-averaged consistency targets improve semi-supervised deep learning results"

Notifications You must be signed in to change notification settings

vertix/mean-teacher

 
 

Repository files navigation

Mean teacher

This is the source code for the ICLR 2017 workshop paper "Mean teachers are better role models: Weight-averaged consistency targets improve semi-supervised deep learning results".

Usage

The code runs on Python 3. Install the dependencies and prepare the datasets with the following commands:

pip install tensorflow numpy scipy
./download_svhn.py

To train the model, run:

  • python train_svhn.py to train on SVHN using 500 labels
  • python train_cifar10.py to train on CIFAR-10 using 4000 labels

To reproduce the results of the poster above, run:

  • python train_svhn_final_eval.py to reproduce the SVHN results
  • python train_cifar10_final_eval.py to reproduce the CIFAR-10 results

About

Source code for "Mean teachers are better role models: Weight-averaged consistency targets improve semi-supervised deep learning results"

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • Python 99.2%
  • Shell 0.8%