diff --git a/实验八/1.get_dataset.py b/实验八/1.get_dataset.py deleted file mode 100644 index 04b8ac1..0000000 --- a/实验八/1.get_dataset.py +++ /dev/null @@ -1,111 +0,0 @@ -import bz2 -import os -import tempfile - -import requests -import tqdm -import tarfile - -# 计算大小 -def get_human_readable_size(size_in_bytes): - # 定义单位 - units = ['B', 'KB', 'MB', 'GB', 'TB', 'PB', 'EB'] - # 计算单位和大小 - index = 0 - while size_in_bytes >= 1024 and index < len(units) - 1: - size_in_bytes /= 1024 - index += 1 - return f"{size_in_bytes:.2f} {units[index]}" - -# 下载与进度条 -def download(url,output_path): - filename = url.split('/')[-1] - download_path = os.path.join(output_path,filename) - - response=requests.get(url,stream=True) - total_size = int(response.headers.get('content-length', 0)) - os.makedirs(output_path,exist_ok=True) - - print("下载 ",filename,' 到 ',output_path,' |大小= ',get_human_readable_size(total_size)) - - # 如已下载，跳过 - if os.path.exists(download_path): - print(filename, ' 已存在，跳过') - return - - with open(download_path,'wb') as file,tqdm.tqdm( - desc='下载进度', - total=total_size, - unit='B', - unit_scale=True, - unit_divisor=1024 - )as bar: - for chunk in response.iter_content(chunk_size=1024): - file.write(chunk) - bar.update(len(chunk)) - - -def decompress(file_path, output_dir): - os.makedirs(output_dir, exist_ok=True) - print('解压 '+file_path.split('/')[-1],' 到 ',output_dir) - with tarfile.open(file_path) as tar: - members = tar.getmembers() - with tqdm.tqdm(total=len(members), desc='解压进度', unit='file') as bar: - for member in members: - # 构建完整路径 - full_path = os.path.join(output_dir, member.name) - - # 因FDDB存在同一路径下存在相同文件的问题，特此跳过 - # 如果文件已存在，跳过解压 - if os.path.exists(full_path): - bar.update(1) - continue - - # 解压文件 - tar.extract(member, path=output_dir) - bar.update(1) - - -def decompress_bz2(file_path, output_dir): - # 确保输出目录存在 - os.makedirs(output_dir, exist_ok=True) - print('解压 ' + file_path.split('/')[-1], ' 到 ', output_dir) - - # 获取.bz2文件的名称，不包含扩展名 - output_filename = os.path.basename(file_path).rsplit('.bz2', 1)[0] - - # 构建输出文件的完整路径 - output_file_path = os.path.join(output_dir, output_filename) - - # 检查输出文件是否已存在，如果存在则跳过解压 - if os.path.exists(output_file_path): - print(output_filename, ' 已存在，跳过解压') - return - - # 解压.bz2文件 - with bz2.BZ2File(file_path, 'rb') as bz2_file, open(output_file_path, 'wb') as output_file: - output_file.write(bz2_file.read()) - -# 下载人脸数据集 -face_dataset_url='http://vis-www.cs.umass.edu/fddb/originalPics.tar.gz' -face_dataset_path='cache/dataset/face/' -download(face_dataset_url,face_dataset_path) - -# 下载人脸数据标签 -face_label_path='cache/dataset/face/label/' -face_label_url='http://vis-www.cs.umass.edu/fddb/FDDB-folds.tgz' -download(face_label_url,face_label_path) - -# 解压数据集 -decompress(os.path.join(face_dataset_path,face_dataset_url.split('/')[-1]),face_dataset_path) -decompress(os.path.join(face_label_path,face_label_url.split('/')[-1]),face_label_path) - -# 下载ResNet模型 -model_url1='https://github.com/davisking/dlib-models/raw/refs/heads/master/shape_predictor_68_face_landmarks.dat.bz2' -model_url2='https://github.com/davisking/dlib-models/raw/refs/heads/master/dlib_face_recognition_resnet_model_v1.dat.bz2' -model_path='models/' -download(model_url1,model_path) -download(model_url2,model_path) - -decompress_bz2(os.path.join(model_path,model_url1.split('/')[-1]),model_path) -decompress_bz2(os.path.join(model_path,model_url2.split('/')[-1]),model_path) \ No newline at end of file diff --git a/实验八/2.pretrain.py b/实验八/2.pretrain.py deleted file mode 100644 index 5004214..0000000 --- a/实验八/2.pretrain.py +++ /dev/null @@ -1,31 +0,0 @@ -import cv2 as cv -import os -import tqdm - -# 处理图像 -def pretrain(img_path,output_path): - img=cv.imread(img_path, cv.IMREAD_GRAYSCALE) - _,img=cv.threshold(img,127,255,cv.THRESH_BINARY) - img=cv.blur(img,(3,3)) - os.makedirs(os.path.dirname(output_path), exist_ok=True) - cv.imwrite(output_path, img) - -# 获取文件路径 -def get_img_name(directory, extensions=None): - if extensions is None: - extensions = ['.png', '.jpg','.jpeg'] - files = [] - for root, dirs, file_names in os.walk(directory): - for file_name in file_names: - if any(file_name.lower().endswith(ext) for ext in extensions): - files.append(os.path.join(root, file_name)) - return files - -# 处理人脸数据集 -os.makedirs('cache/pretrained/face/', exist_ok=True) -face_files=get_img_name('cache/dataset/face/') -print('预处理人脸数据中：') -for img_path in tqdm.tqdm(face_files): - relative_path=os.path.relpath(img_path, 'cache/dataset/face/') - output_path=os.path.join('cache/pretrained/face', relative_path) - pretrain(img_path, output_path) diff --git a/实验八/3.train.py b/实验八/3.train.py deleted file mode 100644 index 09a2734..0000000 --- a/实验八/3.train.py +++ /dev/null @@ -1,74 +0,0 @@ -import dlib -import cv2 as cv -import joblib -import numpy as np -import os - -from sklearn.ensemble import RandomForestClassifier -from sklearn.model_selection import train_test_split -from sklearn.multioutput import MultiOutputClassifier -from tqdm import tqdm - -# 提取人脸特征 -def extract_face_features(image_path): - img = cv.imread(image_path) - detections = detector(img, 1) - face_features = [] - face_bboxes = [] - for rect in detections: - shape = predictor(img, rect) - face_descriptor = face_rec_model.compute_face_descriptor(img, shape) - face_features.append(face_descriptor) - bbox = (rect.left(), rect.top(), rect.right(), rect.bottom()) - face_bboxes.append(bbox) - return face_features, face_bboxes - -# 获取图片路径 -def get_img_path(directory, extension=None): - if extension is None: - extension = ['.jpg', '.jpeg', '.png'] - files = [] - for root, dirs, file_names in os.walk(directory): - for file_name in file_names: - if any(file_name.lower().endswith(ext) for ext in extension): - files.append(os.path.join(root, file_name)) - return files - - -# 加载dlib模型 -detector = dlib.get_frontal_face_detector() -predictor = dlib.shape_predictor('models/shape_predictor_68_face_landmarks.dat') -face_rec_model = dlib.face_recognition_model_v1('models/dlib_face_recognition_resnet_model_v1.dat') - -# 获取图片 -img_directory = 'cache/pretrained/' -images = get_img_path(img_directory) - -# 提取特征 -features = [] -labels = [] - -for image in tqdm(images, desc='提取图片特征中：'): - extracted_features, face_bboxes = extract_face_features(image) - for feature,bbox in zip(extracted_features, face_bboxes): - features.append(feature) - labels.append(bbox) - -X_train = np.array(features) -Y_train = np.array(labels) - -# 分割测试集 -X_train, X_test, Y_train, Y_test = train_test_split(X_train, Y_train, test_size=0.2, random_state=42) - -# 训练SVM模型 -print('训练模型中') -clf=MultiOutputClassifier(RandomForestClassifier(n_estimators=100,random_state=42)) -clf.fit(X_train, Y_train) - -# 评估训练数据 -predictions = clf.predict(X_test) -accuracy=(predictions==Y_test).mean() -print(f'分类器准确度：{accuracy * 100:.2f}%') - -os.makedirs('models/', exist_ok=True) -joblib.dump(clf, 'models/my_classifier.pkl') diff --git a/实验八/detect.py b/实验八/detect.py new file mode 100644 index 0000000..e69de29 diff --git a/实验八/main.py b/实验八/main.py deleted file mode 100644 index eb389a0..0000000 --- a/实验八/main.py +++ /dev/null @@ -1,16 +0,0 @@ -# 这是一个示例 Python 脚本。 - -# 按 Shift+F10 执行或将其替换为您的代码。 -# 按 双击 Shift 在所有地方搜索类、文件、工具窗口、操作和设置。 - - -def print_hi(name): - # 在下面的代码行中使用断点来调试脚本。 - print(f'Hi, {name}') # 按 Ctrl+F8 切换断点。 - - -# 按装订区域中的绿色按钮以运行脚本。 -if __name__ == '__main__': - print_hi('PyCharm') - -# 访问 https://www.jetbrains.com/help/pycharm/ 获取 PyCharm 帮助 diff --git a/实验八/train.py b/实验八/train.py new file mode 100644 index 0000000..d7c8931 --- /dev/null +++ b/实验八/train.py @@ -0,0 +1,95 @@ +import cv2 +import cv2 as cv +import dlib +import joblib +import numpy as np +from sklearn.metrics import accuracy_score +from sklearn.model_selection import train_test_split +from sklearn.svm import SVC +from tqdm import tqdm + +# 方法区 +## 转换为二值图和GRB图 +def img_cvt(img): + img_gray = cv.cvtColor(img, cv2.COLOR_BGR2GRAY) + _,img_gray=cv.threshold(img_gray,127,255,cv.THRESH_BINARY) + img_gray = cv.blur(img_gray,(3,3)) + img_rgb = cv.cvtColor(img, cv.COLOR_BGR2RGB) + return img_gray,img_rgb +## 提取人脸特征 +def extract_face_feature(img): + faces=detector(img) + if len(faces)>0: + face=faces[0] + landmarks=predictor(img,face) + face_descriptor=face_rec_model.compute_face_descriptor(img,landmarks) + return np.array(face_descriptor) + else: + return None +## 生成伪样品 +def gen_negative_samples(features_dim=128): + return np.random.uniform(-1,1,features_dim) + +# 变量区 +dataset_video=cv.VideoCapture('cache/dataset.mp4') +datasets=[] +datasets_rgb=[] +features=[] +negative_samples=[] + +# 收集数据集 +## 采用视频的格式收集图片 +while(True): + ret,frame=dataset_video.read() + if not ret: + break + datasets.append(frame) + +# 加载人脸检测模型 +detector=dlib.get_frontal_face_detector() +predictor=dlib.shape_predictor("models/shape_predictor_68_face_landmarks_GTX.dat") +face_rec_model=dlib.face_recognition_model_v1('models/dlib_face_recognition_resnet_model_v1.dat') + +# 预处理数据集 +for dataset in tqdm(datasets,desc='正在预处理数据集'): + _,dataset_rgb = img_cvt(dataset) + datasets_rgb.append(dataset_rgb) + +# 求人脸特征 +for dataset in tqdm(datasets_rgb,desc='正在求人脸特征'): + face_descriptor=extract_face_feature(dataset) + if face_descriptor is not None: + features.append(face_descriptor) + +# 生成负样本 +for i in tqdm(range(len(features)*15),desc='正在生成负样本'): + negative_sample=gen_negative_samples() + negative_samples.append(negative_sample) + +# 编码数据集 +positive_labels=[1]*len(features) +negative_labels=[0]*len(negative_samples) + +samples=features+negative_samples +labels=positive_labels+negative_labels + +X_train=np.array(samples) +Y_train=np.array(labels) + +# 分离数据集和训练集 +X_train,X_test,Y_train,Y_test=train_test_split(X_train,Y_train,test_size=0.2,random_state=42,stratify=Y_train) + +# 训练 +print('正在训练分类器') +classifier=SVC(kernel='linear',C=1.0,random_state=42) +classifier.fit(X_train,Y_train) + +# 评估分类器性能 +Y_pred=[] +for test_sample in tqdm(X_test,desc='评估分类器'): + Y_pred.append(classifier.predict(test_sample.reshape(1,-1))) +accuracy=accuracy_score(Y_test,Y_pred) +print(f'准确率：{accuracy*100:.2f}%') + +# 保存模型 +joblib.dump(classifier,'models/classifier.pkl') \ No newline at end of file