def showResults(threshold, key, tasking, people):
found = list()
uniq = set()
#for enc in encodings:
#print(type(all_enc),type(sensativity))
for enc in tasking[key]['face_vec']:
for candidate in people:
c_enc = people[candidate]['face_vec']
c_pic = people[candidate]['pic']
c_tim = people[candidate]['times']
dist = face.face_distance([enc], c_enc)
if dist < threshold:
#print ('candidate:{0} detections:{1} dist:{2:0.3}'.format(candidate,len(c_tim),dist[0]))
RGB_img = cv2.cvtColor(c_pic, cv2.COLOR_BGR2RGB)
if candidate not in uniq:
uniq.add(candidate)
found.append((candidate, len(c_tim), dist[0], RGB_img))
side = math.ceil(math.sqrt(len(found)))
plt.figure(figsize=(20, 20))
sorted_by_distance = sorted(found, key=lambda tup: tup[2])
stats = makestats(sorted_by_distance)
for idx, f in enumerate(sorted_by_distance):
plt.subplot(side, side, idx + 1)
plt.title("label:{0} dist:{1:02.3} det:{2:03}".format(
f[0], f[2], f[1]))
plt.axis('off')
plt.imshow(PIL.Image.fromarray(f[3]))
print(stats)
plt.show()
def post(self):
ret_val = {}
additional_vectors = None
loc_enc_h = handle_post_file(additional_vectors)
if len(loc_enc_h) != 2:
abort(404,
message='expected exactly 2 faces, found {0}'.format(
len(loc_enc_h)))
ret_val['meta'] = {}
ret_val['meta']['vector_set'] = {}
ret_val['meta']['vector_set']['count'] = len(loc_enc_h)
ret_val['meta']['vector_set']['vectors'] = []
for loc, enc, h in loc_enc_h:
d = {}
d['hash'] = h
d['face_pic_hash'] = h
d['face_coordinates'] = list(loc)
d['vector'] = list(enc)
ret_val['meta']['vector_set']['vectors'].append(d)
left = np.array(loc_enc_h[0][1])
right = np.array(loc_enc_h[1][1])
ret_val['results'] = {}
ret_val['results']['distance'] = face.face_distance([left], right)[0]
return ret_val
def make_result_array(cls, meta):
distance = meta['query']['threshold']
#print('meta.keys:', meta.keys())
# sys.stdout.flush()
vectors = meta['vector_set']['vectors']
video_array = []
for entity in vectors:
query_src_hash = entity['hash']
query_src_enc = np.array(entity['vector'])
for key in face_pickle:
s_entity = face_pickle[key]
s_vec = np.array(s_entity['face_vec'])
#s_pic = s_entity['face_pic']
s_videos = s_entity['videos']
left = query_src_enc
right = s_vec
vector_distance = face.face_distance([left], right)[0]
if vector_distance < distance:
entry = {}
entry['distance'] = vector_distance
entry['src'] = query_src_hash
entry['uri'] = write_file(s_entity)
entry['hash'] = entry['uri'].split('/')[-1].split('.')[0]
entry['videos'] = cls.proc_videos(s_videos)
video_array.append(entry)
video_array.sort(key=lambda temp_d: temp_d['distance'])
# print('videoArray*********************************')
# print(video_array)
# print('videoArray*********************************')
return video_array
def post(self):
ret_val = {}
dist_name = 'threshold'
if dist_name not in request.form:
abort(404, message='threshold not specified')
distance = float(request.form[dist_name])
if distance > 1.0 or distance < 0:
abort(
404,
message='threshold {0} must be between [0,1]'.format(distance))
if 'vector' not in request.form:
abort(404, message='vector not specified')
if 'video' not in request.form:
abort(404, message='video not specified')
search_vector = request.form['vector']
search_vector = json.loads(search_vector)
video = request.form['video']
ret_val['meta'] = {}
ret_val['meta']['vector_set'] = {}
ret_val['meta']['vector_set']['count'] = 1
ret_val['meta']['distance'] = distance
ret_val['meta']['video'] = video
ret_val['meta']['vector_set']['vectors'] = []
d = {}
d['hash'] = None
d['face_pic_hash'] = None
d['face_coordinates'] = None
d['vector'] = search_vector
ret_val['meta']['vector_set']['vectors'].append(d)
ret_val['results'] = {}
video_hits = {}
for key in face_pickle:
s_entity = face_pickle[key]
s_pic = s_entity['face_pic']
s_vec = np.array(s_entity['face_vec'])
s_videos = s_entity['videos']
left = search_vector
right = s_vec
vector_distance = face.face_distance([left], right)[0]
if vector_distance < distance:
if video in s_videos:
d = {}
d['face_pic'] = 'comming soon'
d['frames'] = []
for frame in s_videos[video]:
d['frames'].append(frame)
video_hits[key] = d
ret_val['results'] = video_hits
ret_val['meta']['num_results'] = len(video_hits.keys())
return ret_val
def post(self):
ret_val = {}
additional_vectors = None
dist_name = 'threshold'
if dist_name not in request.form:
abort(404, message='threshold not specified')
distance = float(request.form[dist_name])
if distance > 1.0 or distance < 0:
abort(
404,
message='threshold {0} must be between [0,1]'.format(distance))
sys.stdout.flush()
loc_enc_h = handle_post_file(additional_vectors)
if len(loc_enc_h) != 1:
abort(404,
message='expected exactly 1 faces, found {0}'.format(
len(loc_enc_h)))
video_hits = defaultdict(int)
query_src_loc, query_src_enc, query_src_h = loc_enc_h[0]
for key in face_pickle:
s_entity = face_pickle[key]
s_vec = np.array(s_entity['face_vec'])
s_videos = s_entity['videos']
left = query_src_enc
right = s_vec
vector_distance = face.face_distance([left], right)[0]
if vector_distance < distance:
for vid_key in s_videos:
video_hits[vid_key] += len(s_videos[vid_key])
ret_val['meta'] = {}
ret_val['meta']['distance'] = distance
ret_val['meta']['vector_set'] = {}
ret_val['meta']['vector_set']['count'] = len(loc_enc_h)
ret_val['meta']['vector_set']['vectors'] = []
for loc, enc, h in loc_enc_h:
d = {}
d['hash'] = h
d['face_pic_hash'] = h
d['face_coordinates'] = list(loc)
d['vector'] = list(enc)
ret_val['meta']['vector_set']['vectors'].append(d)
ret_val['results'] = []
for vh in video_hits:
d = {}
d['video'] = vh
d['count'] = video_hits[vh]
ret_val['results'].append(d)
return ret_val