def feature_extraction_N2N(self,
image_dir,
N1=0,
N2=10000,
template_dir=None,
enhanced_img_path=None):
subject_paths = glob.glob(image_dir + '*')
assert (len(subject_paths) > 0)
if not os.path.exists(template_dir):
os.makedirs(template_dir)
subject_paths = subject_paths[N1:N2]
for subject_path in subject_paths:
img_files = glob.glob(subject_path + '/*.png')
assert (len(img_files) > 0)
img_files.sort()
for i, img_file in enumerate(img_files):
print i, img_file
img_name = os.path.basename(img_file)
fname = template_dir + os.path.splitext(img_name)[0] + '.dat'
if os.path.exists(fname):
continue
rolled_template = self.feature_extraction_single_rolled(
img_file, output_dir=template_dir, ppi=1200)
stop = timeit.default_timer()
if rolled_template is not None:
print(fname)
template.Template2Bin_Byte_TF_C(fname,
rolled_template,
isLatent=True,
save_mask=False)
assert (N2 - N1 > 0)
assert (template_dir is not None)
for i in range(N1, N2 + 1):
start = timeit.default_timer()
img_file = os.path.join(image_dir, str(i) + '.bmp')
img_name = os.path.basename(img_file)
fname = template_dir + os.path.splitext(img_name)[0] + '.dat'
if os.path.exists(fname):
continue
rolled_template = self.feature_extraction_single_rolled(
img_file, output_dir=template_dir)
stop = timeit.default_timer()
if rolled_template is not None:
print(fname)
template.Template2Bin_Byte_TF_C(fname,
rolled_template,
isLatent=True,
save_mask=False)
print stop - start
def feature_extraction_MSP(self,
image_dir,
N1=0,
N2=10000,
template_dir=None,
enhanced_img_path=None):
assert (N2 - N1 > 0)
assert (template_dir is not None)
for i in range(N1, N2 + 1):
start = timeit.default_timer()
img_file = os.path.join(image_dir, str(i) + '.bmp')
img_name = os.path.basename(img_file)
fname = template_dir + os.path.splitext(img_name)[0] + '.dat'
if os.path.exists(fname):
continue
rolled_template = self.feature_extraction_single_rolled(
img_file, output_dir=template_dir)
stop = timeit.default_timer()
if rolled_template is not None:
print(fname)
template.Template2Bin_Byte_TF_C(fname,
rolled_template,
isLatent=True,
save_mask=False)
print stop - start
else:
print("rolled_template is None!")
print stop - start
def feature_extraction(self, image_dir, img_type='bmp', template_dir=None, enhancement=False):
img_files = glob.glob(image_dir + '*.' + img_type)
assert(len(img_files) > 0)
img_files.sort()
for i, img_file in enumerate(img_files):
print img_file
start = timeit.default_timer()
img_name = os.path.basename(img_file)
img_name = os.path.splitext(img_name)[0]
fname = template_dir + img_name + '.dat'
if os.path.exists(fname):
continue
if enhancement:
rolled_template, enhanced_img = self.feature_extraction_single_rolled_enhancement(img_file)
if template_dir is not None:
enhanced_img = np.asarray(enhanced_img, dtype=np.uint8)
io.imsave(os.path.join(template_dir, img_name + '.jpeg'), enhanced_img)
else:
rolled_template = self.feature_extraction_single_rolled(img_file, output_dir=template_dir)
stop = timeit.default_timer()
print stop - start
if template_dir is not None:
fname = template_dir + img_name + '.dat'
print(fname)
template.Template2Bin_Byte_TF_C(fname, rolled_template, isLatent=False)
def feature_extraction_Longitudinal(self,
image_dir,
img_type='bmp',
template_dir=None,
enhancement=False,
N1=0,
N2=10000):
subjects = os.listdir(image_dir)
subjects.sort()
assert (len(subjects) > 16000)
subjects = subjects[N1:N2]
for i, subject in enumerate(subjects):
for finger_ID in range(10):
img_files = glob.glob(
os.path.join(image_dir, subject,
'*' + str(finger_ID) + '.bmp'))
img_files.sort()
if len(img_files) < 5:
continue
img_files = img_files[:5]
for img_file in img_files:
start = timeit.default_timer()
img_name = os.path.basename(img_file)
img_name = os.path.splitext(img_name)[0]
if template_dir is not None:
fname = template_dir + subject + '_' + img_name + '.dat'
if os.path.exists(fname):
continue
if enhancement:
rolled_template, enhanced_img = self.feature_extraction_single_rolled_enhancement(
img_file)
if template_dir is not None:
enhanced_img = np.asarray(enhanced_img,
dtype=np.uint8)
io.imsave(
os.path.join(template_dir, img_name + '.jpeg'),
enhanced_img)
else:
rolled_template = self.feature_extraction_single_rolled(
img_file, output_dir=template_dir)
stop = timeit.default_timer()
print stop - start
if template_dir is not None:
fname = template_dir + subject + '_' + img_name + '.dat'
print(fname)
template.Template2Bin_Byte_TF_C(fname,
rolled_template,
isLatent=False)
def feature_extraction(self,
image_dir,
template_dir=None,
minu_path=None,
N1=0,
N2=258):
img_files = glob.glob(image_dir + '*.bmp')
assert (len(img_files) > 0)
if not os.path.exists(template_dir):
os.makedirs(template_dir)
img_files.sort()
if minu_path is not None:
minu_files = glob.glob(minu_path + '*.txt')
minu_files.sort(key=lambda filename: int(''.join(
filter(str.isdigit, filename))))
for i, img_file in enumerate(img_files):
print i, img_file
img_name = os.path.basename(img_file)
if template_dir is not None:
fname = template_dir + os.path.splitext(img_name)[0] + '.dat'
if os.path.exists(fname):
continue
start = timeit.default_timer()
if minu_path is not None and len(minu_files) > i:
latent_template, texture_template = self.feature_extraction_single_latent(
img_file,
output_dir=template_dir,
show_processes=False,
minu_file=minu_files[i],
show_minutiae=False)
else:
# show_minutiae was true here
latent_template, texture_template = self.feature_extraction_single_latent(
img_file,
output_dir=template_dir,
show_processes=False,
show_minutiae=False)
stop = timeit.default_timer()
print stop - start
fname = template_dir + os.path.splitext(img_name)[0] + '.dat'
template.Template2Bin_Byte_TF_C(fname,
latent_template,
isLatent=True)
def main_single_image(image_file, template_dir):
global config
if not config:
dir_path = os.path.dirname(os.path.dirname(os.path.realpath(__file__)))
with open(dir_path + 'config') as config_file:
config = json.load(config_file)
des_model_dirs = []
patch_types = []
model_dir = config['DescriptorModelPatch2']
des_model_dirs.append(model_dir)
patch_types.append(2)
model_dir = config['DescriptorModelPatch8']
des_model_dirs.append(model_dir)
patch_types.append(8)
model_dir = config['DescriptorModelPatch11']
des_model_dirs.append(model_dir)
patch_types.append(11)
# minutiae extraction model
minu_model_dirs = []
minu_model_dirs.append(config['MinutiaeExtractionModelLatentSTFT'])
minu_model_dirs.append(config['MinutiaeExtractionModel'])
# enhancement model
enhancement_model_dir = config['EnhancementModel']
LF_Latent = FeatureExtraction_Latent(
patch_types=patch_types,
des_model_dirs=des_model_dirs,
enhancement_model_dir=enhancement_model_dir,
minu_model_dirs=minu_model_dirs)
# minu_path = None
if not os.path.exists(template_dir):
os.makedirs(template_dir)
print("Latent query: " + image_file)
print("Starting feature extraction (single latent)...")
latent_template, texture_template = LF_Latent.feature_extraction_single_latent(
image_file,
output_dir=template_dir,
show_processes=False,
minu_file=None,
show_minutiae=False)
fname = template_dir + os.path.splitext(
os.path.basename(image_file))[0] + '.dat'
template.Template2Bin_Byte_TF_C(fname, latent_template, isLatent=True)
return fname