def main():
import numpy as np
from models import vgg19_model
from algorithms import neural_best_buddies as NBBs
from util import util
from util import MLS
vgg19 = vgg19_model.define_Vgg19(opt)
img_nums, images = get_imgs()
for imgnum in tqdm(missing):
print(imgnum)
save_dir = os.path.join(opt.results_dir, str(imgnum))
if os.path.exists(os.path.join(save_dir, 'BtoA.npy')):
continue
try:
print('Working on', imgnum)
source_path = os.path.join(opt.source, f'{base_name}_{imgnum}.png')
A = util.read_image(source_path, opt.imageSize)
B = util.read_image(opt.target, opt.imageSize)
print(A.shape, B.shape)
nbbs = NBBs.sparse_semantic_correspondence(
vgg19, opt.gpu_ids, opt.tau, opt.border_size, save_dir,
opt.k_per_level, opt.k_final, opt.fast)
points = nbbs.run(A, B)
mls = MLS.MLS(v_class=np.int32)
mls.run_MLS_in_folder(root_folder=save_dir)
except Exception as e:
print(e)
with open(os.path.join(save_dir, 'no_correspondence.txt'),
'w') as f:
f.write('')
def main():
import numpy as np
from models import vgg19_model
from algorithms import neural_best_buddies as NBBs
from util import util
from tqdm import tqdm
from util import MLS
vgg19 = vgg19_model.define_Vgg19(opt)
img_nums, images = get_imgs()
for imgnum in tqdm(range(start_imgnum, start_imgnum + N)):
save_dir = os.path.join(opt.results_dir, str(img_nums[imgnum]))
print('Working on', images[imgnum])
try:
source_path = os.path.join(opt.source, images[imgnum])
A = util.read_image(source_path, opt.imageSize)
B = util.read_image(opt.target, opt.imageSize)
nbbs = NBBs.sparse_semantic_correspondence(vgg19, opt.gpu_ids, opt.tau,
opt.border_size, save_dir,
opt.k_per_level, opt.k_final,
opt.fast)
points = nbbs.run(A, B)
mls = MLS.MLS(v_class=np.int32)
mls.run_MLS_in_folder(root_folder=save_dir)
except Exception as e:
print(e)
def run_NBB_and_get_points(opt, nbbs):
A = util.read_image(opt.sourceImg, opt.imageSize)
B = util.read_image(opt.targetImg, opt.imageSize)
points = nbbs.run(A, B)
new_file_name = os.path.basename(opt.sourceImg) + "_" + os.path.basename(
opt.targetImg) + ".txt"
with open(os.path.join(opt.results_dir, new_file_name), 'w') as f:
for item in points:
f.write("%s\n" % item)
f.close()
dist = get_distance(points, A, B)
return dist, points
def post(self):
image_key = read_image(self, "project_image")
p = self.create_project(image_key)
session = get_current_session()
self.create_project_admin(p, session['me_email'])
intercomio_api.events(session['me_email'], event_name='created_project')
mixpanel_api.events(session['me_email'], 'created_project')
self.redirect('/projects/skills?project_id='+str(p.id))
def post(self):
image_key = read_image(self, "project_image")
p = self.create_project(image_key)
session = get_current_session()
self.create_project_admin(p, session['me_email'])
intercomio_api.events(session['me_email'],
event_name='created_project')
mixpanel_api.events(session['me_email'], 'created_project')
self.redirect('/projects/skills?project_id=' + str(p.id))
def run(self):
count = 0
for filepath in self.files:
ext = filepath.split('.')[-1]
h5_obj = h5py.File(filepath, 'r') if ext in ('cxi', 'h5') else None
lcls_data = util.get_lcls_data(filepath) if ext == 'lcls' else None
data_shape = util.get_data_shape(filepath)
for i in range(data_shape[self.dataset][0]):
img = util.read_image(filepath,
frame=i,
h5_obj=h5_obj,
lcls_data=lcls_data,
dataset=self.dataset)['image']
if img is None:
continue
img = img.astype(np.float64)
if count == 0:
img_mean = img
img_sigma = np.zeros_like(img_mean)
else:
img_mean_prev = img_mean.copy()
img_mean += (img - img_mean) / (count + 1.)
img_sigma += (img - img_mean_prev) * (img - img_mean)
count += 1
ratio = count * 100. / self.max_frame
self.update_progress.emit(ratio)
if count == self.max_frame:
break
if count >= self.max_frame:
break
img_sigma = img_sigma / count
img_sigma = np.sqrt(img_sigma)
# write to file
time.sleep(0.1)
if self.output is None:
output = 'output.npz'
else:
output = self.output
dir_ = os.path.dirname(output)
if not os.path.isdir(dir_):
os.mkdir(dir_)
np.savez(output, mean=img_mean, sigma=img_sigma)
import time
import numpy as np
from models import vgg19_model
from algorithms import neural_best_buddies as NBBs
from util import util
from util import MLS
from options.options import Options
opt = Options().parse()
vgg19 = vgg19_model.define_Vgg19(opt)
# save_dir = os.path.join(opt.results_dir, opt.name)
save_dir = os.path.join(opt.results_dir)
print('\n\n---------------------------')
print('Started Neural Best-Buddies')
print('---------------------------')
start_time = time.time()
nbbs = NBBs.sparse_semantic_correspondence(vgg19, opt.gpu_ids, opt.tau,
opt.border_size, save_dir,
opt.k_per_level, opt.k_final,
opt.fast)
A = util.read_image(opt.datarootA, opt.imageSize)
B = util.read_image(opt.datarootB, opt.imageSize)
points = nbbs.run(A, B, opt.datarootA, opt.datarootB)
end_time = time.time()
total_time = end_time - start_time
print('\nFinished after {:04.3f} seconds'.format(total_time))
def post(self):
image_key = read_image(self, "project_image")
self.update_project(image_key)
self.redirect('/member/projects/dashboard')
def post(self):
image_key = read_image(self, "project_image")
self.update_project(image_key)
self.redirect('/member/projects/dashboard')