本文共 33375 字，大约阅读时间需要 111 分钟。
##creat_data.py##实现数据的打包import cv2import tensorflow as tf##dlib 实现抠图import dlib##读取标志信息索引值,不同的索引值对应不同的属性anno_file = "/home/lsy/CelebA/Anno/list_attr_celeba.txt"ff = open(anno_file)anno_info = ff.readlines()attribute_class = anno_info[1].split(" ")print(attribute_class)##Eyeglasses，Male,Young,Smilingidx = 0##打印索引值for i in attribute_class: if i =="Eyeglasses": print("Eyeglasses",idx) elif i=="Male": print("Male",idx) elif i =="Young": print("Young",idx) elif i=="Smiling": print("Smiling",idx) idx += 1##打包数据writer_train = tf.python_io.TFRecordWriter("train.tfrecords")writer_test = tf.python_io.TFRecordWriter("test.tfrecords")##人脸检测器detector = dlib.get_frontal_face_detector()##分析for idx in range(2,anno_info.__len__()): info = anno_info[idx]##有些数据有两个空格，需要替换成一个 attr_val = info.replace(" "," ").split(" ") #print(attr_val.__len__())##属性需要加1 print(attr_val[0]) print(attr_val[16]) print(attr_val[21]) print(attr_val[32]) print(attr_val[40])##下载的数据集地址 im_data = cv2.imread("/home/lsy/CelebA/Img/img_celeba.7z/img_celeba/" + attr_val[0]) rects = detector(im_data,0)##如果没检测到人脸 if len(rects) == 0: continue x1 = rects[0].left() y1 = rects[0].top() x2 = rects[0].right() y2 = rects[0].bottom()##需要将检测到的人脸扣下来 y1 = int(max(y1 - 0.3 * (y2 - y1),0))##可视化 #cv2.rectangle(im_data,(x1,y1),(x2,y2),(255,0,0),2) #cv2.imshow("11",im_data) #cv2.waitKey(0)##数据清洗 if y2 - y1 < 50 or x2 - x1 < 50 or x1 < 0 or y1 < 0: continue im_data = im_data[y1:y2,x1:x2] im_data = cv2.resize(im_data,(128,128)) ex = tf.train.Example( features = tf.train.Features( feature = { "image":tf.train.Feature( bytes_list = tf.train.BytesList(value=[im_data.tobytes()]) ), "Eyeglasses":tf.train.Feature( int64_list = tf.train.Int64List( value = [int(attr_val[16])] ) ), "Male": tf.train.Feature( int64_list=tf.train.Int64List( value=[int(attr_val[21])] ) ), "Young": tf.train.Feature( int64_list=tf.train.Int64List( value=[int(attr_val[32])] ) ), "Smiling": tf.train.Feature( int64_list=tf.train.Int64List( value=[int(attr_val[40])] ) ) } ) )##序列化，生成训练集与测试集 if idx > anno_info.__len__() * 0.95: writer_test.write(ex.SerializeToString()) else: writer_train.write(ex.SerializeToString()) ##搭建模型##train.pyimport tensorflow as tffrom tensorflow.contrib.layers import *from tensorflow.contrib.slim.python.slim.nets.inception_v3 import inception_v3_baseslim = tf.contrib.slim##nets下有很多网络结构,可以自己换def inception_v3 (images,drop_out = 0.5, is_training = True): batch_norm_params ={ "is_training":is_training, "trainable":True, "decay":0.9997, "epsilon":0.00001, "variables_collections":{ "beta":None, "gamma":None, "moving_mean":["moving_vars"], "moving_variance":["moving_var"] } } weights_regularizer = tf.contrib.layers.l2_regularizer(0.00004) with tf.contrib.slim.arg_scope( [tf.contrib.slim.conv2d,tf.contrib.slim.fully_connected], weights_regularizer = weights_regularizer, trainable = True): with tf.contrib.slim.arg_scope( [tf.contrib.slim.conv2d], weights_regularizer = tf.truncated_normal_initializer(stddev=0.1), activation_fn = tf.nn.relu, normalizer_fn = batch_norm, normalizer_params = batch_norm_params): ##传入主干网络 nets,endpoints = inception_v3_base(images) print(nets) print(endpoints) net = tf.reduce_mean(nets,axis=[1,2])##nhwc net = tf.nn.dropout(net,drop_out,name = "droplast") net = flatten(net,scope="flatten")##全连接层 net_eyeglass = slim.fully_connected(net,2,activation_fn = None) net_young = slim.fully_connected(net, 2, activation_fn = None) net_male = slim.fully_connected(net, 2, activation_fn = None) net_smiling = slim.fully_connected(net, 2, activation_fn = None) return net_eyeglass,net_young,net_male,net_smiling##喂数据##输入input_xinput_x = tf.placeholder(tf.float32, shape=[None, 128, 128, 3])label_eyeglasses = tf.placeholder(tf.int64, shape=[None, 1])label_young = tf.placeholder(tf.int64, shape=[None, 1])label_male = tf.placeholder(tf.int64, shape=[None, 1])label_smiling = tf.placeholder(tf.int64, shape=[None, 1])logits_eyeglasses, logits_young, logits_male, logits_smiling = inception_v3(input_x,0.5,True)loss_eyeglasses = tf.losses.sparse_softmax_cross_entropy(labels= label_eyeglasses, logits = logits_eyeglasses)loss_young = tf.losses.sparse_softmax_cross_entropy(labels= label_young, logits = logits_young)loss_male = tf.losses.sparse_softmax_cross_entropy(labels= label_male, logits = logits_male)loss_smiling = tf.losses.sparse_softmax_cross_entropy(labels= label_smiling, logits = logits_smiling)logits_eyeglasses = tf.nn.softmax(logits_eyeglasses)logits_young = tf.nn.softmax(logits_young)logits_male = tf.nn.softmax(logits_male)logits_smiling = tf.nn.softmax(logits_smiling)loss = loss_eyeglasses + loss_young + loss_male +loss_smiling##定义正则化reg_set =tf.get_collection(tf.GraphKeys.REGULARIZATION_LOSSES)l2_loss = tf.add_n(reg_set)##learnglobal_step = tf.Variable(0,trainable=True)lr = tf.train.exponential_decay(0.0001, global_step, decay_steps=1000, decay_rate=0.98, staircase=False)updata_ops = tf.get_collection(tf.GraphKeys.UPDATE_OPS)with tf.control_dependencies(updata_ops): train_op = tf.train.AdamOptimizer(lr).minimize(loss + l2_loss , global_step)##喂数据def get_one_batch(batch_size,type): if type == 0: file_list = tf.gfile.Glob("train.tfrecords") else: file_list = tf.gfile.Glob("test.tfrecords") reader = tf.TFRecordReader()##等loss收敛 file_queue = tf.train.string_input_producer( file_list, num_epochs=None, shuffle=True ) _, se = reader.read(file_queue) if type == 0: batch = tf.train.shuffle_batch([se], batch_size, capacity = batch_size, min_after_dequeue = batch_size // 2) else: batch = tf.train.shuffle_batch([se], batch_size, capacity=batch_size) features = tf.parse_example(batch , features={ "image":tf.FixedLenFeature([],tf.string), "Eyeglasses": tf.FixedLenFeature([1], tf.int64), "Male": tf.FixedLenFeature([1], tf.int64), "Young": tf.FixedLenFeature([1], tf.int64), "Smiling": tf.FixedLenFeature([1], tf.int64), })##数据解析 batch_im = features["image"] batch_eye = (features["Eyeglasses"] + 1) // 2 batch_male = (features["Male"]+ 1) // 2 batch_young = (features["Young"]+ 1) // 2 batch_smiling = (features["Smiling"]+ 1) // 2 batch_im = tf.decode_raw(batch_im,tf.uint8) batch_im = tf.cast(tf.reshape(batch_im, (batch_size, 128, 128, 3)) , tf.float32) return batch_im, batch_eye, batch_male, batch_young, batch_smiling##训练tr_im_batch, tr_label_eye_batch, tr_label_male,\tr_label_young, tr_label_smiling = get_one_batch(64,0)te_im_batch, te_label_eye_batch, te_label_male,\te_label_young, te_label_smiling = get_one_batch(64, 1)######with tf.Session() as session: coord = tf.train.Coordinator() tf.train.start_queue_runners(sess=session,coord=coord) init_ops = tf.group(tf.global_variables_initializer(), tf.local_variables_initializer()) session.run(init_ops) #ckpt = tf.train.get_checkpoint_state("models") #### #if ckpt and tf.train.checkpoint_exists(ckpt.models_checkpoint_path): # saver.restore(session,ckpt.model_checkpoint_path) for step in range(100000): batch_x, batch_eye, batch_male, batch_young, batch_smiling = session.run( [tr_im_batch, tr_label_eye_batch, tr_label_male, tr_label_young, tr_label_smiling]) _,loss_val,l2_loss_val,lr_val,global_step_val,loss_eyeglasses_val,loss_male_val,loss_smiling_val,loss_young_val = \ session.run([train_op, loss, l2_loss, lr, global_step, loss_eyeglasses, loss_male, loss_smiling, loss_young], feed_dict={ input_x:batch_x, label_eyeglasses:batch_eye, label_male:batch_male, label_smiling:batch_smiling, label_young:batch_young }) print("ite:{}, loss:{}, l2_loss:{}, loss_eye:{}, loss_mail:{}, loss_young:{}, loss_smiling:{}, lr:{}".format(step,loss_val,l2_loss_val, loss_eyeglasses_val, loss_male_val, loss_young_val, loss_smiling_val,lr_val))##每一万保存一次模型 #if step % 10000 ==0 : # saver.save(sess = session , save_path = "models/models-{}.ckpt".format(step)) import tensorflow as tfimport cv2import globimport numpy as npimport dlibpb_path = "face_attribute.pb"detector = dlib.get_frontal_face_detector()sess = tf.Session()with sess.as_default(): with tf.gfile.FastGFile(pb_path,'rb') as f : graph_def = sess.graph_def graph_def.ParseFromString(f.read()) tf.import_graph_def(graph_def,name = "")pred_eyeglasses = sess.graph.get_tensor_by_name("Softmax:0")pred_young = sess.graph.get_tensor_by_name("Softmax_1:0")pred_male = sess.graph.get_tensor_by_name("Softmax_2:0")pred_smiling = sess.graph.get_tensor_by_name("Softmax_3:0")im_list = glob.glob("/home/lsy/CelebA/Img/img_celeba.7z/img_celeba/*")for im_path in im_list: im_data = cv2.imread(im_path) #im_data = cv2.imread("/home/lsy/CelebA/Img/img_align_celeba.7z/img_celeba/" + attr_val[0]) rects = detector(im_data,0) if len(rects) == 0: continue x1 = rects[0].left() y1 = rects[0].top() x2 = rects[0].right() y2 = rects[0].bottom() y1 = int(max(y1 - 0.3 * (y2 - y1),0)) im_data = im_data[y1:y2,x1:x2] im_data = cv2.resize(im_data,(128,128)) print(im_data.shape) [eye,young,male,smiling] =sess.run([pred_eyeglasses,pred_young,pred_male,pred_smiling], {"Placeholder:0":np.expand_dims(im_data,0)}) print("eye,young,male,smiling",eye,young,male,smiling) cv2.imshow("111",im_data) cv2.waitKey(0) import tensorflow as tffrom tensorflow.contrib.layers import *from tensorflow.contrib.slim.python.slim.nets.inception_v3 import inception_v3_baseslim = tf.contrib.slimdef inception_v3 (images,drop_out = 0.5, is_training = True): batch_norm_params ={ "is_training":is_training, "trainable":True, "decay": 0.9997, "epsilon": 0.00001, "variables_collections": { "beta": None, "gamma": None, "moving_mean": ["moving_vars"], "moving_variance": ["moving_var"] } } weights_regularizer = tf.contrib.layers.l2_regularizer(0.00004) with tf.contrib.slim.arg_scope( [tf.contrib.slim.conv2d,tf.contrib.slim.fully_connected], weights_regularizer = weights_regularizer, trainable = True): with tf.contrib.slim.arg_scope( [tf.contrib.slim.conv2d], weights_regularizer = tf.truncated_normal_initializer(stddev=0.1), activation_fn = tf.nn.relu, normalizer_fn = batch_norm, normalizer_params = batch_norm_params): nets,endpoints = inception_v3_base(images) print(nets) print(endpoints) net = tf.reduce_name(nets,axis=[1,2]) net = tf.nn.dropout(net,drop_out,name = "droplast") net = flatten(net , scope="flatten") net_eyeglass = slim.fully_connected(net, 2,activation_fn = None) net_young = slim.fully_connected(net, 2, activation_fn=None) net_male = slim.fully_connected(net, 2, activation_fn=None) net_smiling = slim.fully_connected(net,2,activation_fn = None) return net_eyeglass,net_young,net_male,net_smilinginput_x = tf.placeholder(tf.float32,shape=[None,128,128,3])label_eyeglasses = tf.placeholder(tf.int64,shape=[None,1])label_young = tf.placeholder(tf.int64,shape=[None,1])label_male = tf.placeholder(tf.int64,shape=[None,1])label_smiling = tf.placeholder(tf.int64,shape=[None,1])logits_eyeglasses,logits_young,logits_male,logits_smiling = inception_v3(input_x,1.0,False)loss_eyeglasses = tf.losses.sparse_softmax_cross_entropy(labels= label_eyeglasses,logits = logits_eyeglasses)loss_young = tf.losses.sparse_softmax_cross_entropy(labels= label_young,logits = logits_young)loss_male = tf.losses.sparse_softmax_cross_entropy(labels= label_male,logits = logits_male)loss_smiling = tf.losses.sparse_softmax_cross_entropy(labels= label_smiling,logits = logits_smiling)logits_eyeglasses = tf.nn.softmax(logits_eyeglasses)logits_young = tf.nn.softmax(logits_young)logits_male = tf.nn.softmax(logits_male)logits_smiling = tf.nn.softmax(logits_smiling)logits_eyeglasses = tf.argmax(logits_eyeglasses,axis=1)logits_young = tf.argmax(logits_young,axis=1)logits_male = tf.argmax(logits_male,axis=1)logits_smiling = tf.argmax(logits_smiling,axis=1)print(logits_eyeglasses,logits_young,logits_male,logits_smiling)loss = loss_eyeglasses + loss_young + loss_male +loss_smilingreg_set =tf.get_collection(tf.GraphKeys.REGULARIZATION_LOSSES)l2_loss = tf.add_n(reg_set)global_step = tf.Variable(0,trainable=True)lr = tf.train.exponential_decay(0.0001,global_step,decay_steps=1000,staircase=False)updata_ops = tf.get_collection(tf.GraphKeys.UPDATE_OPS)with tf.contrib_dependencies(updata_ops): train_op = tf.train.AdamOptimizer(lr).minimize(loss + l2_loss ,global_step)def get_one_batch(batch_size,type): if type == 0: file_list = tf.gfile.Glob("train.tfrecords") else: file_list = tf.gfile.Glob("test.tfrecords") reader = tf.TFRecordReader() file_queue = tf.train.string_input_producer( file_list,num_epochs=None,shuffle=True ) _, se = reader.read(file_queue) if type == 0: batch = tf.train.shuffle_batch([se], batch_size, capacity= batch_size,min_after_dequeue=batch_size // 2) else: batch = tf.train.shuffle_batch([se], batch_size, capacity=batch_size) features = tf.parse_example(batch,features={ "image":tf.FixedLenFeature([],tf.string), "Eyeglasses": tf.FixedLenFeature([1], tf.int64), "Male": tf.FixedLenFeature([1], tf.int64), "Young": tf.FixedLenFeature([1], tf.int64), "Smiling": tf.FixedLenFeature([1], tf.int64), }) batch_im = features["image"] batch_eye = (features["Eyeglasses"] + 1) // 2 batch_male = (features["Male"]+ 1) // 2 batch_young = (features["Young"]+ 1) // 2 batch_smiling = (features["Smiling"]+ 1) // 2 batch_im = tf.decode_raw(batch_im,tf.uint8) batch_im = tf.cast(tf.reshape(batch_im,(batch_size,128,128,3)),tf.float32) return batch_im,batch_eye,batch_male,batch_young,batch_smiling tr_im_batch,tr_label_eye_batch,tr_label_male,tr_label_young,tr_label_smiling =get_one_batch(64,0) te_im_batch, te_label_eye_batch, te_label_male, te_label_young, te_label_smiling = get_one_batch(64, 1) with tf.Session() as session: coord = tf.train.Coordinator() tf.train.start_queue_runners(sess=session,coord=coord) init_ops = tf.group(tf.global_variables_initializer(),tf.lacal_variables_initializer()) session.run(init_ops) ckpt = tf.train.get_checkpoint_state("models") if ckpt and tf.train.checkpoint_exists(ckpt.models_checkpoint_path): print(ckpt.model_checkpoint_path) saver.restore(session,ckpt.model_checkpoint_path)################################################################################# output_graph_def = tf.graph_util.\ convert_variables_to_constants(session,session.graph.as_graph_def(), ['ArgMax','ArgMax_1','ArgMax_2','ArgMax_3']) with tf.gfile.FastGFile("face_attribute.pb","wb")as f: f.write(output_graph_def.SerializeToString()) f.close() from flask import Flask,requestfrom object_detection.utils import ops as utils_opsimport osimport numpy as npos.environ['TF_CPP_MIN_LOG_LEVEL'] = '2'os.environ["CUDA_DEVICE_ORDER"] = "PCI_BUS_ID"os.environ["CUDA_VISIBLE_DEVICES"] = "0"import cv2from gevent import monkeymonkey.patch_all()import tensorflow as tfimport numpy as npapp = Flask(__name__)PATH_TO_FROZEN_GRAPH = "/home/lsy/flask_server_facerecognition/frozen_inference_graph.pb"PATH_TO_LABELS = "/home/lsy/flask_server_facerecognition/object_detection/face_label_map.pbtxt"IMAGE_SIZE = (256, 256)detection_sess = tf.Session()face_recognition_sess = tf.Session()with detection_sess.as_default(): od_graph_def = tf.GraphDef() with tf.gfile.GFile(PATH_TO_FROZEN_GRAPH, 'rb') as fid: serialized_graph = fid.read() od_graph_def.ParseFromString(serialized_graph) tf.import_graph_def(od_graph_def, name='') ops = tf.get_default_graph().get_operations() all_tensor_names = {output.name for op in ops for output in op.outputs} tensor_dict = {} for key in [ 'num_detections', 'detection_boxes', 'detection_scores', 'detection_classes', 'detection_masks' ]: tensor_name = key + ':0' if tensor_name in all_tensor_names: tensor_dict[key] = tf.get_default_graph().get_tensor_by_name( tensor_name) if 'detection_masks' in tensor_dict: # The following processing is only for single image detection_boxes = tf.squeeze(tensor_dict['detection_boxes'], [0]) detection_masks = tf.squeeze(tensor_dict['detection_masks'], [0]) # Reframe is required to translate mask from box coordinates to image coordinates and fit the image size. real_num_detection = tf.cast(tensor_dict['num_detections'][0], tf.int32) detection_boxes = tf.slice(detection_boxes, [0, 0], [real_num_detection, -1]) detection_masks = tf.slice(detection_masks, [0, 0, 0], [real_num_detection, -1, -1]) detection_masks_reframed = utils_ops.reframe_box_masks_to_image_masks( detection_masks, detection_boxes, IMAGE_SIZE[0], IMAGE_SIZE[1]) detection_masks_reframed = tf.cast( tf.greater(detection_masks_reframed, 0.5), tf.uint8) # Follow the convention by adding back the batch dimension tensor_dict['detection_masks'] = tf.expand_dims( detection_masks_reframed, 0) image_tensor = tf.get_default_graph().get_tensor_by_name('image_tensor:0')#####################################face featureface_feature_sess = tf.Session()ff_pb_path = "face_recognition_model.pb"with face_feature_sess.as_default(): ff_od_graph_def = tf.GraphDef() with tf.gfile.GFile(ff_pb_path, 'rb') as fid: serialized_graph = fid.read() od_graph_def.ParseFromString(serialized_graph) tf.import_graph_def(od_graph_def, name='') ff_images_placeholder = tf.get_default_graph().get_tensor_by_name("input:0") ff_train_placeholder = tf.get_default_graph().get_tensor_by_name("phase_train:0") ff_embeddings = tf.get_default_graph().get_tensor_by_name("embeddings:0")####################################人脸属性face_attribute_sess = tf.Session()ff_pb_path = "/home/lsy/flask_server_facerecognition/face_recognition_model.pb"#####with face_attribute_sess.as_default(): ff_attri_graph_def = tf.GraphDef() with tf.gfile.GFile(ff_pb_path, 'rb') as fid: serialized_graph = fid.read() od_graph_def.ParseFromString(serialized_graph) tf.import_graph_def(od_graph_def, name='') pred_eyeglasses = tf.get_default_graph().get_tensor_by_name("ArgMax:0") pred_young = tf.get_default_graph().get_tensor_by_name("ArgMax_1:0") pred_male = tf.get_default_graph().get_tensor_by_name("ArgMax_2:0") pred_smiling = tf.get_default_graph().get_tensor_by_name("ArgMax_3:0") face_attribute_image_tensor = tf.get_default_graph().get_tensor_by_name("Placeholder:0")###face featureface_landmark_sess = tf.Session()ff_pb_path = "/home/lsy/facial_landmark/landmark.pb"with face_landmark_sess.as_default(): ff_od_graph_def = tf.GraphDef() with tf.gfile.GFile(ff_pb_path, 'rb') as fid: serialized_graph = fid.read() od_graph_def.ParseFromString(serialized_graph) tf.import_graph_def(od_graph_def, name='') landmark_tensor = tf.get_default_graph().get_tensor_by_name("fully_connected_1/Relu:0")########@app.route('/face_attribute', methods=['POST', 'GET'])def face_attribute(): f = request.files.get("file") upload_path = os.path.join("tmp/tmp_attribute." + f.filename.split(",")[-1]) f.save(upload_path) print(upload_path) im_data = cv2.imread(upload_path) rects = detector(im_data,0) if len(rects) == 0: ##return "error" x1 = rects[0].left() y1 = rects[0].top() x2 = rects[0].right() y2 = rects[0].bottom() y1 = int(max(y1 - 0.3 * (y2 - y1),0)) im_data = im_data[y1:y2,x1:x2] im_data = cv2.resize(im_data,(128,128)) [eye_glass,young,male,smiling] = face_attribute_sess.run( [pred_eyeglasses,pred_young,pred_male,pred_smiling], feed_dict={face_attribute_image_tensor: np.expand_dims(im_data,0)} ) return "{},{},{},{}".format(eye_glass[0],young[0],male[0],smiling[0])###############import dlibpredictor = dlib.shape_predictor('/home/lsy/facial_landmark/shape_predictor_68_face_landmarks.dat')detector = dlib.get_frontal_face_detector()@app.route('/face_landmark', methods=['POST', 'GET'])def face_landmark(): ##实现图片上传 f = request.files.get('file') print(f) upload_path = os.path.join("tmp/tmp_landmark." + f.filename.split(".")[-1]) # secure_filename(f.filename)) #注意：没有的文件夹一定要先创建，不然会提示没有该路径 print(upload_path) f.save(upload_path) ##人脸检测 im_data = cv2.imread(upload_path) sp = im_data.shape ## im_data = cv2.resize(im_data, IMAGE_SIZE) output_dict = detection_sess.run(tensor_dict, feed_dict={image_tensor: np.expand_dims( im_data, 0)}) # all outputs are float32 numpy arrays, so convert types as appropriate output_dict['num_detections'] = int(output_dict['num_detections'][0]) output_dict['detection_classes'] = output_dict[ 'detection_classes'][0].astype(np.uint8) output_dict['detection_boxes'] = output_dict['detection_boxes'][0] output_dict['detection_scores'] = output_dict['detection_scores'][0] x1 = 0 y1 = 0 x2 = 0 y2 = 0 for i in range(len(output_dict['detection_scores'])): if output_dict['detection_scores'][i] > 0.1: bbox = output_dict['detection_boxes'][i] y1 = bbox[0] x1 = bbox[1] y2 = (bbox[2]) x2 = (bbox[3]) print(output_dict['detection_scores'][i], x1, y1, x2, y2) ##提取人脸区域 y1 = int((y1+(y2 - y1)*0.2) * sp[0]) x1 = int(x1 * sp[1]) y2 = int(y2 * sp[0]) x2 = int(x2 * sp[1]) face_data = im_data[y1:y2, x1:x2] cv2.imwrite("face_landmark.jpg",face_data) face_data = cv2.resize(face_data,(128,128)) pred = face_landmark_sess.run(landmark_tensor, {"Placeholder:0": np.expand_dims(face_data, 0)}) pred = pred[0] res = [] for i in range(0,136,2): res.append(str((pred[i] * (x2-x1) + x1) / sp[1])) res.append(str((pred[i+1] * (y2 - y1) + y1) / sp[0])) res = ",".join(res) return res return "error"@app.route('/face_landmark_tf', methods=['POST', 'GET'])def face_landmark_tf(): f = request.files.get("file") upload_path = os.path.join("tmp/tmp_landmark."+f.filename.split(",")[-1]) f.save(upload_path) ## print(upload_path) ## im_data = cv2.imread(upload_path) im_data = cv2.cvtColor(im_data,cv2.COLOR_BGR2GRAY) sp = im_data.shape rects = detector(im_data,0) res = [] for face in rects: shape = predictor(im_data,face) for pt in shape.parts(): pt_pos = (pt.x,pt.y) x1 = face.left() y1 = face.top() x2 = face.right() y2 = face.bottom() ptx = (pt.x - x1 ) * 1.0 / (x2 - x1) pty = (pt.y - y1 ) * 1.0 / (y2 - y1) res.append(str(ptx)) res.append(str(pty)) res.append(str(pt.x * 1.0 / sp[1])) res.append(str(pt.y * 1.0/ sp[0])) if res.__len__() == 136 * 2: res = ",".join(res) print(res) return res return "error"@app.route("/")def helloword(): return 'Hello World!
'@app.route('/upload', methods=['POST', 'GET'])def upload(): f = request.files.get('file') print(f) upload_path = os.path.join("tmp/tmp." + f.filename.split(".")[-1]) # secure_filename(f.filename)) #注意：没有的文件夹一定要先创建，不然会提示没有该路径 print(upload_path) f.save(upload_path) return upload_path@app.route("/face_detect")def inference(): im_url = request.args.get("url") im_data = cv2.imread(im_url) sp = im_data.shape im_data = cv2.resize(im_data, IMAGE_SIZE) output_dict = detection_sess.run(tensor_dict, feed_dict={image_tensor: np.expand_dims( im_data, 0)}) # all outputs are float32 numpy arrays, so convert types as appropriate output_dict['num_detections'] = int(output_dict['num_detections'][0]) output_dict['detection_classes'] = output_dict[ 'detection_classes'][0].astype(np.uint8) output_dict['detection_boxes'] = output_dict['detection_boxes'][0] output_dict['detection_scores'] = output_dict['detection_scores'][0] x1 = 0 y1 = 0 x2 = 0 y2 = 0 for i in range(len(output_dict['detection_scores'])): if output_dict['detection_scores'][i] > 0.1: bbox = output_dict['detection_boxes'][i] y1 = bbox[0] x1 = bbox[1] y2 = (bbox[2]) x2 = (bbox[3]) print(output_dict['detection_scores'][i], x1, y1, x2, y2) return str([x1, y1, x2, y2])# 图像数据标准化def prewhiten(x): mean = np.mean(x) std = np.std(x) std_adj = np.maximum(std, 1.0/np.sqrt(x.size)) y = np.multiply(np.subtract(x, mean), 1/std_adj) return ydef read_image(path): ### im_data = cv2.imread(path) im_data = prewhiten(im_data) im_data = cv2.resize(im_data, (160, 160)) #1 * h * w * 3 return im_data@app.route("/face_feature")def face_feature(): im_data1 = read_image("/home/lsy/64_CASIA-FaceV5/" "crop_image_160/0/0_0000.jpg") im_data1 = np.expand_dims(im_data1, axis=0) emb1 = face_feature_sess.run(ff_embeddings, feed_dict={ff_images_placeholder:im_data1, ff_train_placeholder:False}) strr = ",".join(str(i) for i in emb1[0]) return strr@app.route('/face_register', methods=['POST', 'GET'])def face_register(): ##实现图片上传 f = request.files.get('file') print(f) upload_path = os.path.join("tmp/tmp." + f.filename.split(".")[-1]) # secure_filename(f.filename)) #注意：没有的文件夹一定要先创建，不然会提示没有该路径 print(upload_path) f.save(upload_path) ##人脸检测 im_data = cv2.imread(upload_path) sp = im_data.shape im_data = cv2.resize(im_data, IMAGE_SIZE) output_dict = detection_sess.run(tensor_dict, feed_dict={image_tensor: np.expand_dims( im_data, 0)}) # all outputs are float32 numpy arrays, so convert types as appropriate output_dict['num_detections'] = int(output_dict['num_detections'][0]) output_dict['detection_classes'] = output_dict[ 'detection_classes'][0].astype(np.uint8) output_dict['detection_boxes'] = output_dict['detection_boxes'][0] output_dict['detection_scores'] = output_dict['detection_scores'][0] x1 = 0 y1 = 0 x2 = 0 y2 = 0 for i in range(len(output_dict['detection_scores'])): if output_dict['detection_scores'][i] > 0.1: bbox = output_dict['detection_boxes'][i] y1 = bbox[0] x1 = bbox[1] y2 = (bbox[2]) x2 = (bbox[3]) print(output_dict['detection_scores'][i], x1, y1, x2, y2) ##提取人脸区域 y1 = int(y1 * sp[0]) x1 = int(x1 * sp[1]) y2 = int(y2 * sp[0]) x2 = int(x2 * sp[1]) face_data = im_data[y1:y2, x1:x2] im_data = prewhiten(face_data) #预处理 im_data = cv2.resize(im_data, (160, 160)) im_data1 = np.expand_dims(im_data, axis=0) ##人脸特征提取 emb1 = face_feature_sess.run(ff_embeddings, feed_dict={ff_images_placeholder: im_data1, ff_train_placeholder: False}) strr = ",".join(str(i) for i in emb1[0]) ##写去txt with open("face/feature.txt", "w") as f: f.writelines(strr) f.close() mess = "success" break else: mess = "fail" return mess@app.route('/face_login', methods=['POST', 'GET'])def face_login(): #图片上传 #人脸检测 #人脸特征提取 #加载注册人脸（人脸签到，人脸数很多，加载注册人脸放在face_login, # 启动服务加载/采用搜索引擎/ES） #同注册人脸相似性度量 #返回度量结果 f = request.files.get('file') print(f) upload_path = os.path.join("tmp/login_tmp." + f.filename.split(".")[-1]) # secure_filename(f.filename)) #注意：没有的文件夹一定要先创建，不然会提示没有该路径 print(upload_path) f.save(upload_path) ##人脸检测 im_data = cv2.imread(upload_path) sp = im_data.shape im_data = cv2.resize(im_data, IMAGE_SIZE) output_dict = detection_sess.run(tensor_dict, feed_dict={image_tensor: np.expand_dims( im_data, 0)}) # all outputs are float32 numpy arrays, so convert types as appropriate output_dict['num_detections'] = int(output_dict['num_detections'][0]) output_dict['detection_classes'] = output_dict[ 'detection_classes'][0].astype(np.uint8) output_dict['detection_boxes'] = output_dict['detection_boxes'][0] output_dict['detection_scores'] = output_dict['detection_scores'][0] x1 = 0 y1 = 0 x2 = 0 y2 = 0 for i in range(len(output_dict['detection_scores'])): if output_dict['detection_scores'][i] > 0.1: bbox = output_dict['detection_boxes'][i] y1 = bbox[0] x1 = bbox[1] y2 = (bbox[2]) x2 = (bbox[3]) print(output_dict['detection_scores'][i], x1, y1, x2, y2) ##提取人脸区域 y1 = int(y1 * sp[0]) x1 = int(x1 * sp[1]) y2 = int(y2 * sp[0]) x2 = int(x2 * sp[1]) face_data = im_data[y1:y2, x1:x2] cv2.imwrite("face.jpg",face_data ) im_data = prewhiten(face_data) # 预处理 im_data = cv2.resize(im_data, (160, 160)) im_data1 = np.expand_dims(im_data, axis=0) ##人脸特征提取 emb1 = face_feature_sess.run(ff_embeddings, feed_dict={ff_images_placeholder: im_data1, ff_train_placeholder: False}) with open("face/feature.txt") as f: fea_str = f.readlines() f.close() emb2_str = fea_str[0].split(",") emb2 = [] for ss in emb2_str: emb2.append(float(ss)) emb2 = np.array(emb2) dist = np.linalg.norm(emb1 - emb2) print("dist---->", dist) if dist < 0.3: return "success" else: return "fail" return "fail"@app.route("/face_dis")def face_dis(): im_data1 = read_image("/home/lsy/flask_server_facedetection/tmp/crop_face.jpg") im_data1 = np.expand_dims(im_data1, axis=0) emb1 = face_feature_sess.run(ff_embeddings, feed_dict={ff_images_placeholder:im_data1, ff_train_placeholder:False}) im_data1 = read_image("tmp/face.jpg") im_data1 = np.expand_dims(im_data1, axis=0) emb2 = face_feature_sess.run(ff_embeddings, feed_dict={ff_images_placeholder:im_data1, ff_train_placeholder:False}) dist = np.linalg.norm(emb1 - emb2) return str(dist)if __name__ == '__main__': app.run(host="192.168.0.1, port=90, debug=True)##IP填自己的，用于与前端交互的 还没完全弄好，等我更新哟！！
转载地址：http://pfde.baihongyu.com/