keras实现简单性别识别(二分类问题)
第一步:准备好需要的库
- tensorflow 1.4.0
- h5py 2.7.0
- hdf5 1.8.15.1
- Keras 2.0.8
- opencv-python 3.3.0
- numpy 1.13.3+mkl
- 所需要的人脸检测模块
- mtcnn和opencv
- https://pan.baidu.com/s/1rhP7mcnAtiojhk8eiLroEw
第二步:准备数据集:
将性别不同的图片按照不同的分类放到不同的文件夹内。
数据集
https://pan.baidu.com/s/1_f36Gw4PWztUXZWH_jLWcw
1 import shutil 2 3 # 读取文件中图片信息根据性别分类图片到对应目录中 4 dirroot = "D:\\Users\\a\\Pictures\\adience" 5 f = open(dirroot+"\\fold_frontal_3_data.txt","r") 6 i = 0 7 8 for line in f.readlines(): 9 line = line.split() 10 dir = line[0] 11 12 imgName = "landmark_aligned_face."+ line[2] +'.'+ line[1] 13 if i > 0: 14 if line[5]== "f": 15 print("female") 16 shutil.copy(dirroot+'\\faces\\'+dir+'\\'+imgName, "D:\\pycode\\learn\\data\\validation\\"+imgName) 17 # 移动图片到female目录 18 elif line[5]=="m": 19 print("male") 20 shutil.copy(dirroot+'\\faces\\'+dir+'\\'+imgName, "D:\\pycode\\learn\\data\\validation\\"+imgName) 21 # 移动图片到male目录 22 else: 23 print("N") 24 # 未识别男女 25 i += 1 26 f.close()
使用ImageDataGenerator,来对图片进行归一化和随机旋转。使用flow_from_directory,来自动产生图片标签生成器。
1 class Dataset(object): 2 3 def __init__(self): 4 self.train = None 5 self.valid = None 6 7 8 def read(self, img_rows=IMAGE_SIZE, img_cols=IMAGE_SIZE): 9 train_datagen = ImageDataGenerator( 10 rescale=1. / 255, 11 horizontal_flip=True) 12 13 test_datagen = ImageDataGenerator(rescale=1. / 255) 14 15 train_generator = train_datagen.flow_from_directory( 16 train_data_dir, 17 target_size=(img_rows, img_cols), 18 batch_size=batch_size, 19 class_mode='binary') 20 21 validation_generator = test_datagen.flow_from_directory( 22 validation_data_dir, 23 target_size=(img_rows, img_cols), 24 batch_size=batch_size, 25 class_mode='binary') 26 27 self.train = train_generator 28 self.valid = validation_generator
第三部:网络
1 class Model(object): 2 3 4 5 def __init__(self): 6 self.model = Sequential() 7 self.model.add(Conv2D(32, (3, 3), input_shape=(IMAGE_SIZE,IMAGE_SIZE,3))) 8 self.model.add(Activation('relu')) 9 self.model.add(MaxPooling2D(pool_size=(2, 2))) 10 11 self.model.add(Conv2D(32, (3, 3))) 12 self.model.add(Activation('relu')) 13 self.model.add(MaxPooling2D(pool_size=(2, 2))) 14 15 self.model.add(Conv2D(64, (3, 3))) 16 self.model.add(Activation('relu')) 17 self.model.add(MaxPooling2D(pool_size=(2, 2))) 18 19 self.model.add(Conv2D(64, (3, 3))) 20 self.model.add(Activation('relu')) 21 self.model.add(MaxPooling2D(pool_size=(2, 2))) 22 23 self.model.add(Flatten()) 24 self.model.add(Dense(64)) 25 self.model.add(Activation('relu')) 26 self.model.add(Dropout(0.85)) 27 self.model.add(Dense(1)) 28 self.model.add(Activation('sigmoid')) 29 30 31 def train(self, dataset, batch_size=batch_size, nb_epoch=epochs): 32 33 self.model.compile(loss='binary_crossentropy', 34 optimizer='adam', 35 metrics=['accuracy']) 36 self.model.fit_generator(dataset.train, 37 steps_per_epoch=nb_train_samples // batch_size, 38 epochs=epochs, 39 validation_data=dataset.valid, 40 validation_steps=nb_validation_samples//batch_size) 41 42 43 def save(self, file_path=FILE_PATH): 44 print('Model Saved.') 45 self.model.save_weights(file_path) 46 47 def load(self, file_path=FILE_PATH): 48 print('Model Loaded.') 49 self.model.load_weights(file_path) 50 51 def predict(self, image): 52 # 预测样本分类 53 img = image.resize((1, IMAGE_SIZE, IMAGE_SIZE, 3)) 54 img = image.astype('float32') 55 img /= 255 56 57 #归一化 58 result = self.model.predict(img) 59 print(result) 60 # 概率 61 result = self.model.predict_classes(img) 62 print(result) 63 # 0/1 64 65 return result[0] 66 67 def evaluate(self, dataset): 68 # 测试样本准确率 69 score = self.model.evaluate_generator(dataset.valid,steps=2) 70 print("样本准确率%s: %.2f%%" % (self.model.metrics_names[1], score[1] * 100))
第四部:主程序
1 if __name__ == '__main__': 2 dataset = Dataset() 3 dataset.read() 4 5 6 model = Model() 7 model.load() 8 model.train(dataset) 9 model.evaluate(dataset) 10 model.save()
第五步:识别程序
opencv检测模块版
1 #!/usr/bin/env python 2 """ 3 从摄像头中获取图像实时监测 4 """ 5 import numpy as np 6 import cv2 7 from GenderTrain import Model 8 9 10 def detect(img, cascade): 11 """ 12 检测图像是否含有人脸部分 13 :param img: 待检测帧图像 14 :param cascade: 面部对象检测器 15 :return: 面部图像标记 16 """ 17 rects = cascade.detectMultiScale(img, scaleFactor=1.3, minNeighbors=4, minSize=(30, 30), 18 flags=cv2.CASCADE_SCALE_IMAGE) 19 if len(rects) == 0: 20 return [] 21 rects[:,2:] += rects[:,:2] 22 return rects 23 24 25 def draw_rects(img, rects, color): 26 """ 27 根据图像标记人脸区域与性别 28 :param img: 29 :param rects: 30 :param color: 31 :return: 32 """ 33 for x, y, w, h in rects: 34 face = img[x:x+w,y:y+h] 35 face = cv2.resize(face,(224,224)) 36 if gender.predict(face)==1: 37 text = "Male" 38 else: 39 text = "Female" 40 cv2.rectangle(img, (x, y), (w, h), color, 2) 41 cv2.putText(img, text, (x, h), cv2.FONT_HERSHEY_SIMPLEX, 2.0, (255, 255, 255), lineType=cv2.LINE_AA) 42 43 44 if __name__ == '__main__': 45 haar__cascade_path = "D:\\opencv\\sources\\data\\haarcascades\\haarcascade_frontalface_default.xml" 46 47 cascade = cv2.CascadeClassifier( haar__cascade_path) 48 cam = cv2.VideoCapture(0) 49 # 获取摄像头视频 50 gender = Model() 51 gender.load() 52 # 加载性别模型 53 while True: 54 ret, img = cam.read() 55 # 读取帧图像 56 rects = detect(img, cascade) 57 print(rects) 58 vis = img.copy() 59 draw_rects(vis, rects, (0, 255, 0)) 60 cv2.imshow('Gender', vis) 61 if cv2.waitKey(5) == 27: 62 break 63 cv2.destroyAllWindows()
MTCNN检测版
""" 从摄像头中获取图像实时监测 """ import PIL import numpy as np import detect_face import tensorflow as tf import cv2 from GenderTrain import Model with tf.Graph().as_default(): gpu_options = tf.GPUOptions(per_process_gpu_memory_fraction=0.5) sess = tf.Session(config=tf.ConfigProto(gpu_options=gpu_options, log_device_placement=False)) with sess.as_default(): pnet, rnet, onet = detect_face.create_mtcnn(sess, 'E:\\pycode\\real-time-deep-face-recognition-master\\20170512-110547') minsize = 20 # minimum size of face threshold = [0.6, 0.7, 0.7] # three steps's threshold factor = 0.709 # scale factor margin = 44 frame_interval = 3 batch_size = 1000 image_size = 182 input_image_size = 160 def draw_rects(img, rects, color): """ 根据图像标记人脸区域与性别 :param img: :param rects: :param color: :return: """ for x, y, w, h in rects: face = img[x:x+w,y:y+h] face = cv2.resize(face,(224,224)) if gender.predict(face)==1: text = "Male" else: text = "Female" cv2.rectangle(img, (x, y), (w, h), color, 2) cv2.putText(img, text, (x, h), cv2.FONT_HERSHEY_SIMPLEX, 2.0, (255, 255, 255), lineType=cv2.LINE_AA) if __name__ == '__main__': cam = cv2.VideoCapture(0) # 获取摄像头视频 gender = Model() gender.load() # 加载性别模型 while True: ret, img = cam.read() # 读取帧图像 bounding_boxes, _ = detect_face.detect_face(img, minsize, pnet, rnet, onet, threshold, factor) # 读取帧图像 for face_position in bounding_boxes: face_position = face_position.astype(int) print(face_position[0:4]) rects = [[face_position[0], face_position[1], face_position[2], face_position[3]]] vis = img.copy() draw_rects(vis, rects, (255, 255, 255)) cv2.imshow('Gender', vis) if cv2.waitKey(5) == 27: break cv2.destroyAllWindows()
完全版
import os import random import cv2 import numpy as np from tensorflow.contrib.keras.api.keras.preprocessing.image import ImageDataGenerator,img_to_array from tensorflow.contrib.keras.api.keras.models import Sequential from tensorflow.contrib.keras.api.keras.layers import Dense, Dropout, Activation, Flatten from tensorflow.contrib.keras.api.keras.layers import Conv2D, MaxPooling2D from tensorflow.contrib.keras.api.keras.optimizers import SGD IMAGE_SIZE = 182 # 训练图片大小 epochs = 150#原来是50 # 遍历次数 batch_size = 32 # 批量大小 nb_train_samples = 512*2 # 训练样本总数 nb_validation_samples = 128*2 # 测试样本总数 train_data_dir = 'D:\\code\\learn\\data_sex\\train_data\\' validation_data_dir = 'D:\\data_sex\\test_data\\' # 样本图片所在路径 FILE_PATH = 'Gender_new.h5' # 模型存放路径 class Dataset(object): def __init__(self): self.train = None self.valid = None def read(self, img_rows=IMAGE_SIZE, img_cols=IMAGE_SIZE): train_datagen = ImageDataGenerator( rescale=1. / 255, horizontal_flip=True) test_datagen = ImageDataGenerator(rescale=1. / 255) train_generator = train_datagen.flow_from_directory( train_data_dir, target_size=(img_rows, img_cols), batch_size=batch_size, class_mode='binary') validation_generator = test_datagen.flow_from_directory( validation_data_dir, target_size=(img_rows, img_cols), batch_size=batch_size, class_mode='binary') self.train = train_generator self.valid = validation_generator class Model(object): def __init__(self): self.model = Sequential() self.model.add(Conv2D(32, (3, 3), input_shape=(IMAGE_SIZE,IMAGE_SIZE,3))) self.model.add(Activation('relu')) self.model.add(MaxPooling2D(pool_size=(2, 2))) self.model.add(Conv2D(32, (3, 3))) self.model.add(Activation('relu')) self.model.add(MaxPooling2D(pool_size=(2, 2))) self.model.add(Conv2D(64, (3, 3))) self.model.add(Activation('relu')) self.model.add(MaxPooling2D(pool_size=(2, 2))) self.model.add(Flatten()) self.model.add(Dense(64)) self.model.add(Activation('relu')) self.model.add(Dropout(0.5)) self.model.add(Dense(1)) self.model.add(Activation('sigmoid')) def train(self, dataset, batch_size=batch_size, nb_epoch=epochs): self.model.compile(loss='binary_crossentropy', optimizer='adam', metrics=['accuracy']) self.model.fit_generator(dataset.train, steps_per_epoch=nb_train_samples // batch_size, epochs=epochs, validation_data=dataset.valid, validation_steps=nb_validation_samples//batch_size) def save(self, file_path=FILE_PATH): print('Model Saved.') self.model.save_weights(file_path) def load(self, file_path=FILE_PATH): print('Model Loaded.') self.model.load_weights(file_path) def predict(self, image): # 预测样本分类 img = image.resize((1, IMAGE_SIZE, IMAGE_SIZE, 3)) img = image.astype('float32') img /= 255 #归一化 result = self.model.predict(img) print(result) # 概率 result = self.model.predict_classes(img) print(result) # 0/1 return result[0] def evaluate(self, dataset): # 测试样本准确率 score = self.model.evaluate_generator(dataset.valid,steps=2) print("样本准确率%s: %.2f%%" % (self.model.metrics_names[1], score[1] * 100)) if __name__ == '__main__': dataset = Dataset() dataset.read() model = Model() model.load() model.train(dataset) model.evaluate(dataset) model.save()
本站文章如无特殊说明,均为本站原创,如若转载,请注明出处:keras实现简单性别识别(二分类问题) - Python技术站