def echo(self, mat):
'''
'''
request = pt.matrix_np2proto(mat)
# Run the computation on the server
dist_mat = self.service_stub.echo(request, None)
return pt.matrix_proto2np(dist_mat)
def score(self, score_request):
'''Compare templates to produce scores.'''
score_type = self.scoreType()
result = geo.Matrix()
# Check that this is a known score type
if score_type not in [fsd.SERVER]:
raise NotImplementedError("Score type <%s> not implemented." %
(score_type, ))
# Check to make sure the probe and gallery records are correct
if len(score_request.template_probes.templates) == 0:
raise ValueError("no probe templates were found in the arguments.")
if len(score_request.template_gallery.templates) == 0:
raise ValueError(
"no gallery templates were found in the arguments.")
#THIS IS NOT NECESSAY AS WE ARE ALWAYS COPYING THE TEMPLATES AND NOT USING FACE RECORD -> REFER TO
#FUNCTION in FaceClient.py
'''
if min(len(score_request.face_probes.face_records),len(score_request.template_probes.templates)) != 0:
raise ValueError("probes argument cannot have both face_probes and template_probes defined.")
if max(len(score_request.face_probes.face_records),len(score_request.template_probes.templates)) == 0:
raise ValueError("no probe templates were found in the arguments.")
if min(len(score_request.face_gallery.face_records),len(score_request.template_gallery.templates)) != 0:
raise ValueError("gallery argument cannot have both face_gallery and template_gallery defined.")
if max(len(score_request.face_gallery.face_records),len(score_request.template_gallery.templates)) == 0:
raise ValueError("no gallery templates were found in the arguments.")
'''
#This is the first attempt at computing similarity scores. This is definitely not the fastest approach.
#Also , this is going to be restricted by memory. The whole similarity matrix be be held in memory.
#So for large datasets this might pose a problem
if score_type == fsd.SERVER:
#rows = probe images
#cols = gallery images
sim_mat = np.zeros((len(score_request.template_probes.templates),
len(score_request.template_gallery.templates)),
dtype=np.float32)
roc_probe_template = roc.roc_template()
roc_gallery_template = roc.roc_template()
#roc_gallery_template_array = roc.new_roc_template_array(len(score_request.template_gallery.templates))
sm_metric = roc.new_roc_similarity()
for p in range(0, len(score_request.template_probes.templates)):
self._rocUnFlatten(
score_request.template_probes.templates[p].buffer,
roc_probe_template)
#print roc_probe_template
for g in range(0,
len(score_request.template_gallery.templates)):
#print(p,g)
#if p == 0:
# roc_gallery_template = roc.roc_template()
# self._rocUnFlatten(score_request.template_gallery.templates[g].buffer,roc_gallery_template)
# roc.roc_template_array_setitem(roc_gallery_template_array,g,roc_gallery_template)
#roc_gallery_template = roc.roc_template()
self._rocUnFlatten(
score_request.template_gallery.templates[g].buffer,
roc_gallery_template)
#roc.roc_compare_templates(roc_probe_template, roc.roc_template_array_getitem(roc_gallery_template_array,g), sm_metric)
roc.roc_compare_templates(roc_probe_template,
roc_gallery_template, sm_metric)
sim_mat[p, g] = roc.roc_similarity_value(sm_metric)
#roc.roc_free_template(roc_gallery_template)
roc.delete_roc_similarity(sm_metric)
roc.roc_free_template(roc_probe_template)
roc.roc_free_template(roc_gallery_template)
#for i in range(len(score_request.template_gallery.templates)):
#print(i)
# roc.roc_ensure(roc.roc_free_template(roc.roc_template_array_getitem(roc_gallery_template_array, i)))
else:
NotImplementedError("ScoreType %s is not implemented." %
(score_type, ))
#RankOne returns a similarity score of -1 if it compares with an invalid template
#Threfore find all -1's in the matrix and replace it with a 0
sim_mat[sim_mat == -1.0] = 0.0
#converting the simialrity matrix to distance matrix by subtracting with 1
dist_mat = 1.0 - sim_mat
# Return the result
return pt.matrix_np2proto(dist_mat)
def score(self, score_request):
'''Compare templates to produce scores.'''
score_type = self.scoreType()
result = geo.Matrix()
# Check that this is a known score type
if score_type not in [fsd.L1, fsd.L2, fsd.NEG_DOT]:
raise NotImplementedError("Score type <%s> not implemented." %
(score_type, ))
# Check to make sure the probe and gallery records are correct
if min(len(score_request.face_probes.face_records),
len(score_request.template_probes.templates)) != 0:
raise ValueError(
"probes argument cannot have both face_probes and template_probes defined."
)
if max(len(score_request.face_probes.face_records),
len(score_request.template_probes.templates)) == 0:
raise ValueError("no probe templates were found in the arguments.")
if min(len(score_request.face_gallery.face_records),
len(score_request.template_gallery.templates)) != 0:
raise ValueError(
"gallery argument cannot have both face_gallery and template_gallery defined."
)
if max(len(score_request.face_gallery.face_records),
len(score_request.template_gallery.templates)) == 0:
raise ValueError(
"no gallery templates were found in the arguments.")
# Generate probe and gallery matrices
if len(score_request.face_probes.face_records) > len(
score_request.template_probes.templates):
probe_mat = [
pt.vector_proto2np(face_rec.template.data)
for face_rec in score_request.face_probes.face_records
]
else:
probe_mat = [
pt.vector_proto2np(template.data)
for template in score_request.template_probes.templates
]
probe_mat = np.array(probe_mat, dtype=np.float32)
if len(score_request.face_gallery.face_records) > len(
score_request.template_gallery.templates):
gal_mat = [
pt.vector_proto2np(face_rec.template.data)
for face_rec in score_request.face_gallery.face_records
]
else:
gal_mat = [
pt.vector_proto2np(template.data)
for template in score_request.template_gallery.templates
]
gal_mat = np.array(gal_mat, dtype=np.float32)
# Compute the distance
if score_type == fsd.L1:
dist_mat = spat.distance_matrix(probe_mat, gal_mat, 1)
elif score_type == fsd.L2:
dist_mat = spat.distance_matrix(probe_mat, gal_mat, 2)
elif score_type == fsd.NEG_DOT:
dist_mat = -np.dot(probe_mat, gal_mat.T)
else:
NotImplementedError("ScoreType %s is not implemented." %
(score_type, ))
# Return the result
return pt.matrix_np2proto(dist_mat)
def detect(self, img, face_records, options):
'''Run a face detector and return rectangles.'''
#print('Running Face Detector For ArchFace')
img = img[:, :, ::
-1] #convert from rgb to bgr . There is a reordering from bgr to RGB internally in the detector code.
faces = self.app.get(img)
#print('Number of detections ', dets.shape[0])
# Now process each face we found and add a face to the records list.
idx = -1
for face in faces:
idx += 1
face_record = face_records.face_records.add()
face_record.detection.score = face.det_score
ulx, uly, lrx, lry = face.bbox
#create_square_bbox = np.amax(abs(lrx-ulx) , abs(lry-uly))
face_record.detection.location.CopyFrom(
pt.rect_val2proto(ulx, uly, abs(lrx - ulx), abs(lry - uly)))
face_record.detection.detection_id = idx
face_record.detection.detection_class = "FACE"
#lmark = face_record.landmarks.add()
for ldx in range(0, face.kps.shape[0]):
lmark = face_record.landmarks.add()
lmark.landmark_id = "point_%02d" % ldx
lmark.location.x = face.kps[ldx][0]
lmark.location.y = face.kps[ldx][1]
normed_embedding = face.normed_embedding
face_record.template.data.CopyFrom(
pt.vector_np2proto(normed_embedding))
if hasattr(face, 'landmark_2d_106'):
attribute = face_record.attributes.add()
attribute.key = 'landmark_2d_106'
attribute.matrix.CopyFrom(
pt.matrix_np2proto(face.landmark_2d_106))
if hasattr(face, 'landmark_3d_68'):
attribute = face_record.attributes.add()
attribute.key = 'landmark_3d_68'
attribute.matrix.CopyFrom(
pt.matrix_np2proto(face.landmark_3d_68))
if hasattr(face, 'sex'):
attribute = face_record.attributes.add()
attribute.key = 'sex'
attribute.text = face.sex
if hasattr(face, 'age'):
attribute = face_record.attributes.add()
attribute.key = 'age'
attribute.fvalue = face.age
if options.best:
face_records.face_records.sort(key=lambda x: -x.detection.score)
while len(face_records.face_records) > 1:
del face_records.face_records[-1]