def enroll_face(file, person_id=None, auto_id=False, force=False):
try:
filename = os.path.basename(file)
# Check for duplicates
if database.exists(file) and not force:
print "File '%s' already enrolled" % filename
return
# Make sure we have an identifier
if not person_id and not auto_id:
print "Auto person identification disabled and no identification specified."
return
# File not yet enrolled
print "Processing %s" % filename
# Read data file
absfile = AbsFile(file)
# Derrive filename
if auto_id:
basename = os.path.basename(file)
person_id = basename[:basename.index('d')]
# Create Face object
face = Face(absfile)
# Apply algorithms to process raw data
# Apply selected algorithms
algorithms.process(face, parameters["N"], parameters["K"])
# Compress data
face.compress()
except:
print "File '%s' failed" % file
# In debug mode, show exceptions
if debug:
raise
else:
return
# Enroll to database
database.save(file, person_id, face)
def enroll_face(file, person_id=None, auto_id=False, force=False):
filename = os.path.basename(file)
# Check for duplicates
if database.exists(file) and not force:
print "File '%s' already enrolled" % filename
return
# Make sure we have an identifier
if not person_id and not auto_id:
print "Auto person identification disabled and no identification specified."
return
# File not yet enrolled
print "Processing %s" % filename
# Read data file
absfile = AbsFile(file)
# Derrive filename
if auto_id:
basename = os.path.basename(file)
person_id = basename[:basename.index('d')]
# Create Face object
face = Face(absfile)
# Apply algorithms to process raw data
try:
# Apply selected algorithms
algorithms.process(face, parameters["N"], parameters["K"])
# Compress data
face.compress()
except:
print "File '%s' failed" % file
# In debug mode, show exceptions
if debug:
raise
else:
return
# Enroll to database
database.save(file, person_id, face)
def Run(arguments):
global database, parameters
success = False
result = False
# Initialize a database
database = Database(arguments.database)
# Enrollment
if arguments.enroll:
success = True
is_directory, path = arguments.enroll
if is_directory:
mask = "%s/*.abs"
if arguments.mask:
mask = "%s/*" + arguments.mask + "*.abs"
files = glob.glob(mask % path)
thread_count = 16
chunks = [files[i::thread_count] for i in range(thread_count)]
threads = []
# Process each thread
for chunk in chunks:
thread = threading.Thread(target=lambda x: [
enroll_face(c, arguments.person_id, arguments.auto_id)
for c in x
],
args=(chunk, ))
thread.daemon = True
threads.append(thread)
thread.start()
# Wait for the threads to finish
for thread in threads:
thread.join()
else:
enroll_face(path.name, arguments.person_id, arguments.auto_id)
# Caches for
faces = False
matrix = False
rocs = False
if arguments.copy_to_database and arguments.mask:
database2 = Database(arguments.copy_to_database)
faces_to_copy = [
f for f in list(database.iterator()) if arguments.mask in f[0]
]
for file_name, person_id, face in faces_to_copy:
if database.exists(file_name):
print "File '%s' already exists" % file_name
continue
database2.save(file_name, person_id, face)
# Authenticating
if arguments.authenticate:
print "Authenticating face from '%s'" % arguments.authenticate
success = True
# Create face from file
face = Face(AbsFile(arguments.authenticate))
# Normalize it
algorithms.process(face, parameters["N"], parameters["K"])
# Get the other data
if not faces:
faces = list(database.iterator())
matrix = algorithms.similarity_matrix([face] +
[face[2] for face in faces],
limit=1) # One line matrix
# Evaluate result
methods, _, _ = matrix.shape
tresholds = [0.00, 0.00, 0.00]
result = 3 * [None]
for i in range(methods):
# Select indexes of candidates
vector = numpy.array(matrix[i][0][1:])
candidates, = numpy.where(vector >= tresholds[i])
persons = {}
# Verify candidates
if len(candidates) == 0:
#print "Method %d does not yield any candidates!" % i
continue
# Print method
#print "Results for method %d:" % i
# Print each candidate
for candidate in candidates:
filename, person_id, data = faces[candidate]
# Add person to list of persons
if person_id not in persons:
persons[person_id] = []
persons[person_id].append(matrix[i][0][candidate + 1])
result[i] = [(person, ["%.2f" % s for s in scores])
for person, scores in persons.iteritems()]
# Print results
# for person, scores in persons.iteritems():
# print "Match with person %s with scores %s" % (person, ["%.2f" % s for s in scores])
# Reevaluation
if arguments.reevaluate:
print "Reevaluate faces"
success = True
# Get data
if not faces: faces = list(database.iterator())
# Action
[
algorithms.features_histogram(face[2], parameters["N"],
parameters["K"]) for face in faces
]
# Visualizing
if arguments.depth_map:
print "Generating depth map"
success = True
# Get data
if not faces: faces = list(database.iterator())
# Action
utils.generate_depth_map(faces, arguments.depth_map,
arguments.draw_key_points)
if arguments.feature_map:
print "Generating feature map"
success = True
# Get data
if not faces: faces = list(database.iterator())
# Action
utils.generate_feature_map(faces, arguments.feature_map)
if arguments.similarity_matrix:
print "Generating similarity matrix"
success = True
# Get data
if not faces: faces = list(database.iterator())
if not matrix:
matrix = algorithms.similarity_matrix([face[2] for face in faces])
# Action
utils.generate_similarity_matrix(matrix, faces,
arguments.similarity_matrix)
if arguments.roc_curve:
print "Generating ROC curve"
success = True
# Get data
if not faces: faces = [f for f in list(database.iterator()) if f]
if not matrix:
matrix = algorithms.similarity_matrix([face[2] for face in faces])
if not rocs:
rocs = algorithms.calculate_roc_eer(matrix,
[face[1] for face in faces])
utils.generate_roc_curve(rocs, arguments.roc_curve)
return success, result
def main():
global database, parameters
success = False
# Parse arguments
arguments, parser = argument_parser()
# Parse parameters
try:
for parameter in arguments.parameters.split(','):
key, value = parameter.split('=', 1)
parameters[key] = value
except:
print "Invalid parameters: %s" % arguments.parameters
sys.exit(1)
# Make sure parameters are of right type
for key, value in parameters.iteritems():
try:
parameters[key] = parameter_types[key](value)
except:
print "Parameter '%s' of incorrect type" % key
sys.exit(1)
# Initialize a database
database = Database(arguments.database)
# Enrollment
if arguments.enroll:
success = True
is_directory, path = arguments.enroll
if is_directory:
files = glob.glob("%s/*.abs" % path)
thread_count = 16
chunks = [ files[i::thread_count] for i in range(thread_count) ]
threads = []
# Process each thread
for chunk in chunks:
thread = threading.Thread(target=lambda x: [ enroll_face(c, arguments.person_id, arguments.auto_id) for c in x ], args=(chunk, ))
thread.daemon = True
threads.append(thread)
thread.start()
# Wait for the threads to finish
for thread in threads:
thread.join()
else:
enroll_face(path, arguments.person_id, arguments.auto_id)
# Caches for
faces = False
matrix = False
rocs = False
# Authenticating
if arguments.authenticate:
print "Authenticating face from '%s'" % arguments.authenticate
success = True
# Create face from file
face = Face(AbsFile(arguments.authenticate))
# Normalize it
algorithms.process(face, parameters["N"], parameters["K"])
# Get the other data
if not faces: faces = list(database.iterator())
matrix = algorithms.similarity_matrix([face] + [ face[2] for face in faces ], limit=1) # One line matrix
# Evaluate result
methods, _, _ = matrix.shape
tresholds = [0.90, 0.90, 0.90]
for i in range(methods):
# Select indexes of candidates
vector = numpy.array(matrix[i][0][1:])
candidates, = numpy.where(vector >= tresholds[i])
persons = {}
# Verify candidates
if len(candidates) == 0:
print "Method %d does not yield any candidates!" % i
continue
# Print method
print "Results for method %d:" % i
# Print each candidate
for candidate in candidates:
if candidate == 0:
continue
filename, person_id, data = faces[candidate]
# Add person to list of persons
if person_id not in persons:
persons[person_id] = []
persons[person_id].append(matrix[i][0][candidate + 1])
# Print results
for person, scores in persons.iteritems():
print "Match with person %s with scores %s" % (person, [ "%.2f" % s for s in scores ])
# Reevaluation
if arguments.reevaluate:
print "Reevaluate faces"
success = True
# Get data
if not faces: faces = list(database.iterator())
# Action
[ algorithms.features_histogram(face[2], parameters["N"], parameters["K"]) for face in faces ]
# Visualizing
if arguments.depth_map:
print "Generating depth map"
success = True
# Get data
if not faces: faces = list(database.iterator())
# Action
utils.generate_depth_map(faces, arguments.depth_map, arguments.draw_key_points)
if arguments.feature_map:
print "Generating feature map"
success = True
# Get data
if not faces: faces = list(database.iterator())
# Action
utils.generate_feature_map(faces, arguments.feature_map)
if arguments.similarity_matrix:
print "Generating similarity matrix"
success = True
# Get data
if not faces: faces = list(database.iterator())
if not matrix: matrix = algorithms.similarity_matrix([ face[2] for face in faces ])
# Action
utils.generate_similarity_matrix(matrix, faces, arguments.similarity_matrix)
if arguments.roc_curve:
print "Generating ROC curve"
success = True
# Get data
if not faces: faces = list(database.iterator())
if not matrix: matrix = algorithms.similarity_matrix([ face[2] for face in faces ])
if not rocs: rocs = algorithms.calculate_roc_eer(matrix, [ face[1] for face in faces ])
utils.generate_roc_curve(rocs, arguments.roc_curve)
# Print help in case of no action
if not success:
parser.print_help()
sys.exit(1)
else:
sys.exit(0)
