def main():
if request.method == 'POST':
imageFile = request.files['file']
file_bytes = np.asarray(bytearray(imageFile.read()), dtype=np.uint8)
image = cv2.imdecode(file_bytes, cv2.IMREAD_COLOR)
uniqueCode = hashlib.sha1(image).hexdigest()[:16]
basepath = "/home/ubuntu/face-search-engine"
directory = os.path.join(basepath, 'static/images', str(uniqueCode))
fileUniqueName = str(uniqueCode) + '.jpg'
os.system("mkdir -p %s" % directory)
cv2.imwrite(os.path.join(directory, fileUniqueName), image)
# extract the faces
face = fdv.Face(os.path.join(directory, fileUniqueName), directory)
face.detect_face(detector=detector)
face.write_clip(write_to_file=True)
face.vectorize(sess=sess,
model=model,
model_input_size=(IMAGE_SZ, IMAGE_SZ))
input_faces = [str(i) + ".png" for i in range(face.num_faces)]
output_faces = {}
origin_img = {}
scores = {}
count = 0
for feature in face.face_features:
tmp = client.search(feature.reshape(512), K=10)
out = []
org_url = []
score = []
for f in tmp:
key = f['key']
score.append(round(f['score'], 3))
out.append(key + '.png')
org_url.append(f['meta'])
scores[input_faces[count]] = score
output_faces[input_faces[count]] = out
origin_img[input_faces[count]] = org_url
count += 1
columns = ["Similar Face %s" % (i + 1) for i in range(10)]
return render_template('analyze.html',
directUniqueName=uniqueCode,
fileUniqueName=fileUniqueName,
column_names=columns,
input_faces=input_faces,
output_faces=output_faces,
output_scores=scores,
origin_img=origin_img)
return render_template('index.html')
def analyze():
if request.method == 'POST':
#f = request.form['image']
#user_input = request.form.get("name")
# current path
basepath = os.path.dirname(__file__)
print(basepath)
path = os.path.join(basepath, 'static/images', 'test.jpg')
face = fdv.Face(path, os.path.join(basepath, 'static/images/'), 30)
face.detect_face()
face.write_clip(write_to_file=True)
face.vectorize(sess=sess,
model=model,
model_input_size=(IMAGE_SZ, IMAGE_SZ))
input_faces = [str(i) + ".png" for i in range(face.num_faces)]
output_faces = {}
origin_img = {}
scores = {}
count = 0
for feature in face.face_features:
tmp = client.search(feature.reshape(512), K=5)
out = []
org = []
score = []
for f in tmp:
barcode = f['key'].split('_')
score.append(round(f['score'], 5))
out.append(barcode[0] + '/' + barcode[1] + '.png')
org.append(barcode[0] + '/origin.png')
scores[input_faces[count]] = score
output_faces[input_faces[count]] = out
origin_img[input_faces[count]] = org
count += 1
return render_template('analyze.html',
input_faces=input_faces,
output_faces=output_faces,
output_scores=scores,
origin_img=origin_img) #, userinput=user_input)
return render_template('upload.html')
def process(barcode, url):
# Construct directory structure
save_dir = os.path.join('data', barcode[4:6], barcode[6:8], barcode)
try:
os.makedirs(save_dir)
except:
if os.path.exists(os.path.join(save_dir, 'info')):
logf.write("%s aleardy exist\n" % save_dir)
return
pass
try:
face = fdv.Face(image_path=url,
output_dir=save_dir,
is_url=True,
key=barcode)
face.detect_face(detector=detector)
face.write_clip(write_to_file=True)
face.write_info("info")
logf.write("%s done\n" % save_dir)
except Exception as e:
print(e)
os.system("mv %s %s_fail" % (save_dir, save_dir))
logf.write("Failed to process %s (%s): %s\n" % (url, barcode, e))
import tensorflow as tf
import numpy as np
import face_detect_vectorize as fdv
IMAGE_SZ = 160
MODEL_PATH = '20180402-114759/model-20180402-114759'
client = fawn.Fawn('http://127.0.0.1:8000')
model = fdv.Model(path=MODEL_PATH, image_size=IMAGE_SZ)
config = tf.ConfigProto()
config.gpu_options.allow_growth = True
print("Load faceNet =>>>")
with tf.Session(config=config) as sess:
sess.run(tf.global_variables_initializer())
sess.run(tf.local_variables_initializer())
model.loader(sess)
print("Done!\n")
face = fdv.Face('sample_input.png', None, 30)
face.detect_face()
face.write_clip(write_to_file=False)
face.vectorize(sess=sess,
model=model,
model_input_size=(IMAGE_SZ, IMAGE_SZ))
for feature in face.face_features:
res = client.search(feature.reshape(512), K=5)
print(res)
print("-----------")
with tf.Session(config=config) as sess:
sess.run(tf.global_variables_initializer())
sess.run(tf.local_variables_initializer())
model.loader(sess)
print("Done!\n")
for image in ORIGINAL_IMAGE:
print('Processing %s ... ' % image)
faces_info = []
target_dir = IMAGE_OUTPUT_PATH + os.path.splitext(
os.path.basename(image))[0]
print("Save to %s" % target_dir)
if not os.path.exists(target_dir):
os.system('mkdir -p ' + target_dir)
try:
face = fdv.Face(image_path=image,
output_dir=target_dir,
min_face_size=MIN_FACE_SIZE)
face.detect_face()
face.write_origin()
face.write_clip(write_to_file=True)
face.write_label()
face.vectorize(sess=sess,
model=model,
model_input_size=(IMAGE_SZ, IMAGE_SZ))
face.write_info("info")
except:
os.system('mv %s %s_fail' % (target_dir, target_dir))
sys.exit()
IMAGE_SZ = 160
MODEL_PATH = '20180402-114759/model-20180402-114759'
client = fawn.Fawn('http://127.0.0.1:8888')
SAMPLE = sys.argv[1]
model = fdv.Model(path=MODEL_PATH, image_size=IMAGE_SZ)
config = tf.ConfigProto()
config.gpu_options.allow_growth = True
print("Load faceNet =>>>")
with tf.Session(config=config) as sess:
sess.run(tf.global_variables_initializer())
sess.run(tf.local_variables_initializer())
model.loader(sess)
print("Done!\n")
detector = Detector()
detector.set_min_face_size(30)
face = fdv.Face(SAMPLE, None)
face.detect_face(detector=detector)
face.write_clip(write_to_file=False)
face.vectorize(sess=sess,
model=model,
model_input_size=(IMAGE_SZ, IMAGE_SZ))
for feature in face.face_features:
res = client.search(feature.reshape(512), K=5)
print(res)
print("-----------")