Precision and Recall GAN Evaluations using Manifolds
Implementation of the metrics from this paper:
"Improved Precision and Recall Metric for Assessing Generative Models" https://arxiv.org/pdf/1904.06991v1.pdf
In Keras.
*Note: This is an unofficial implementation. It is also less efficient, using CPU, and takes O(n^2) time. Feel free to contribute!
- Place data in /data/ folder.
- Import get_prec_and_recall function, or add to precision-recall.py.
- Enter the folder names of the real and fake data respectively as parameters to the function.
- Optionally: Set n, meaning number of samples to use.
- Optionally: Set get_scores, if True this will return realism scores for fake samples.
- Run function.