MVM+ : Generative Model via Adversarial Metric Learning

Follow-up study of "Manifold Matching via Deep Metric Learning for Generative Modeling" (ICCV 2021).

Origin: https://github.com/dzld00/pytorch-manifold-matching

Paper: https://arxiv.org/abs/2106.10777

Objective functions

Objective for metric learning:

metric_loss = Cos_(ml_real_out,ml_real_out_shuffle,ml_fake_out_shuffle) 

Objective for manifold matching with learned metric:

g_loss = d_hausdorff_(ml_real_out, ml_fake_out) 

Dataset

Download data to the data path. The sample code uses CelebA dataset.

Training

To train a model for unconditonal generation, run:

python train.py
python train_MNIST.py
python train_synthetic.py

       

Citation

@misc{daiandhang2021manifold,
      title={Manifold Matching via Deep Metric Learning for Generative Modeling}, 
      author={Mengyu Dai and Haibin Hang},
      year={2021},
      eprint={2106.10777},
      archivePrefix={arXiv}
}