leonardo-lyh / blg-piaa

BLG-PIAA

#Pytorch implementation of "Personalized Image Aesthetics Assessment via Meta-Learning With Bilevel Gradient Optimization"

• PyTorch 0.4.1 (with Python 3.6.0) implementation of the following paper.

• If you find our work is useful, pleaes cite our paper:
  @ARTICLE{Zhu2020Personalized,
  author={H. {Zhu} and L. {Li} and J. {Wu} and S. {Zhao} and G. {Ding} and Generic. {Shi}},
  journal={IEEE Trans. Cybern.},
  title={Personalized Image Aesthetics Assessment via Meta-learning with Bilevel Gradient Optimization},
  year={2020},
  note={doi:10.1109/TCYB.2020.2984670}
  }

• finetune_model_newtask.py for personalized aesthetic model fine-tuning.

• train_meta_model.py for aesthetic prior model training.

• If you want to train the aesthetic prior model, you need download FLICKR-AES and AADB databases.