search

wuhuikai / TF-A2RL

The official implementation for A2-RL: Aesthetics Aware Rinforcement Learning for Automatic Image Cropping
MIT License
191 stars 41 forks source link

How to train the model from scratch? #5

Closed w121211 closed 6 years ago

w121211 commented 6 years ago

Thank you for sharing your work. It helps a lot. It seems the code does not include training part? (please correct me if I'm wrong) Would you kindly include it as well, thanks so much.

wuhuikai commented 6 years ago

We couldn't distribute our training code currently mainly for two reasons:

  1. We have the following work going on besides the work itself is under review.
  2. Our training code is based on pytorch, which is some of messy.

You could refer to https://github.com/wuhuikai/a3c-deepmind-pytorch with some modification.

nishiyasu commented 5 years ago

Thank you for sharing your work. I read your papers and got interested in your work. Though your code works fine, I want to train from scratch. So, would you kindly share your training code ? Thank you.

wuhuikai commented 5 years ago

Please contact the 1st author Debang Li for the training code.