def main():
chainer.cuda.get_device_from_id(args.gpu).use()
np.random.seed(1234)
xp = gen.xp
n = int(5000 * args.splits)
#for _ in range(50):
# gen(128)
print("Gen")
ims = gen_images(gen, n, batchsize=125).astype("f")
print(np.max(ims), np.min(ims))
if args.tf:
# mean, std = inception_score.get_inception_score(ims, args.splits)
stat = np.load(args.FID_stat_file, allow_pickle=False)
is_mean, is_std, fid_mean, fid_std = inception_score_tf.get_inception_and_FID(ims, args.splits, ref_stats=stat)
print(is_mean, is_std, fid_mean, fid_std)
else:
model = load_inception_model(args.inception_model_path)
mean, std = inception_score(model, ims, splits=args.splits)
print(mean, std)
if not os.path.exists(args.results_dir):
os.makedirs(args.results_dir)
np.savetxt('{}/inception_score.txt'.format(args.results_dir),
np.array([is_mean, is_std]))
np.savetxt('{}/FID.txt'.format(args.results_dir),
np.array([fid_mean, fid_std]))
def evaluation(trainer=None):
model = load_inception_model(path)
ims = gen_images(gen, n_ims, batchsize=batchsize).astype("f")
mean, std = inception_score(model, ims, splits=splits)
chainer.reporter.report({'inception_mean': mean, 'inception_std': std})
if dst is not None:
preview_dir = '{}/stats'.format(dst)
preview_path = preview_dir + '/inception_score_{:0>8}.txt'.format(
trainer.updater.iteration if trainer is not None else None)
np.savetxt(preview_path, np.array([mean, std]))
def evaluation(trainer=None):
model = load_inception_model(path)
ims = gen_images(gen, n_ims, batchsize=batchsize).astype("f")
mean, std = inception_score(model, ims, splits=splits)
chainer.reporter.report({'inception_mean': mean, 'inception_std': std})
if dst is not None:
preview_dir = '{}/IS.txt'.format(dst)
with open(preview_dir, 'a', encoding='ascii') as f:
f.write(str(trainer.updater.iteration))
f.write(':')
f.write(str(mean))
f.write(', ')
f.write(str(std))
f.write('\n')
def evaluation(trainer=None):
model = load_inception_model(path)
ims = gen_images(gen, n_ims, batchsize=batchsize).astype("f")
mean, std = inception_score(model, ims, splits=splits)
chainer.reporter.report({
'inception_mean': mean,
'inception_std': std
})
if dst is not None:
preview_dir = '{}/stats'.format(dst)
preview_path = preview_dir + '/inception_score_{:0>8}.txt'.format(
trainer.updater.iteration if trainer is not None else None)
np.savetxt(preview_path, np.array([mean, std]))
# # export the best model.
global best_inception
if trainer is not None and best_inception < mean:
ext = extensions.snapshot_object(gen, gen.__class__.__name__ + '_best_inception.npz')
ext(trainer)
best_inception = mean