Code for Birthday Paradox test for diversity of GANs, proposed in Do GANs actually learn the distribution? An empirical study
based on DCGAN-Tensorflow
- Python, Numpy, Tensorflow
Download the model checkpoint and unzip it
$ wget https://www.dropbox.com/s/sfdoyvfl5eozfa5/checkpoint.zip
$ unzip checkpoint.zip
Generate a pool of
$ python main.py --topK=20 --num_batches=6 --sample_dir=samples
The top 20 potential collisions will be in ./samples
import numpy as np
import heapq as hq
import copy as cp
images = sample_from_gan() # a batch of generated samples
queue = [] # a priority queue maintaining top K most similar pairs
n_image = images.shape[0]
topK = 20 # keep top
for i in range(n_image):
for j in range(i+1, n_image):
# measure similarity in pixel space (could be done in some embedding space too)
dist = np.sum((images[i] - images[j])**2)
if len(queue) == 0 or -1*dist > queue[0][0]:
hq.heappush(queue, (-1*cp.deepcopy(dist), cp.deepcopy(images[i]), cp.deepcopy(images[j])))
if len(queue) > topK:
hq.heappop(queue)
for idx in range(topK):
neg_dist, img1, img2 = hq.heappop(queue)
scipy.misc.imsave(config.sample_dir + '/pair#%d_%f_%d.png'%(idx, -1*neg_dist, 1), (img1+1.)/2)
scipy.misc.imsave(config.sample_dir + '/pair#%d_%f_%d.png'%(idx, -1*neg_dist, 2), (img2+1.)/2)To estimate the diversity of a GAN model:
- Sample a batch of size
sfrom the generator - Use the code above to select topK (say 20) potential collisions
- Visually inspect the flagged pairs and check for duplicates
- Repeat
If this test reveals that samples of size