Datasets:
Download lazy1/predict.py from Beifengdo/Multimedia-Technology: direct link, hf CLI and curl.
- Browser
- Download file 4.03 kB
-
https://huggingface.co/datasets/Beifengdo/Multimedia-Technology/resolve/main/lazy1/predict.py
- Command line
-
hf download hf://datasets/Beifengdo/Multimedia-Technology/lazy1/predict.py
-
curl -L -o predict.py https://huggingface.co/datasets/Beifengdo/Multimedia-Technology/resolve/main/lazy1/predict.py
4.03 kB
| import tensorflow.keras as keras | |
| import cv2 | |
| IMAGE_SIZE=64 | |
| #裁剪照片得到64*64的尺寸 | |
| def resize_image(image, height=IMAGE_SIZE, width=IMAGE_SIZE): | |
| top, bottom, left, right = (0, 0, 0, 0) | |
| # 获取图像尺寸 | |
| h, w, _ = image.shape | |
| # 对于长宽不相等的图片,找到最长的一边 | |
| longest_edge = max(h, w) | |
| # 计算短边需要增加多上像素宽度使其与长边等长 | |
| if h < longest_edge: | |
| dh = longest_edge - h | |
| top = dh // 2 | |
| bottom = dh - top | |
| elif w < longest_edge: | |
| dw = longest_edge - w | |
| left = dw // 2 | |
| right = dw - left | |
| else: | |
| pass | |
| # RGB颜色 | |
| BLACK = [0, 0, 0] | |
| # 给图像增加边界,是图片长、宽等长,cv2.BORDER_CONSTANT指定边界颜色由value指定 | |
| constant = cv2.copyMakeBorder(image, top, bottom, left, right, cv2.BORDER_CONSTANT, value=BLACK) | |
| # 调整图像大小并返回 | |
| return cv2.resize(constant, (height, width)) | |
| ##########预测函数######################## | |
| def face_predict(image,model): | |
| image = resize_image(image) | |
| image = image.reshape((1, IMAGE_SIZE, IMAGE_SIZE, 3)) | |
| image=image.astype('float32') | |
| image/=255 | |
| # 给出输入属于各个类别的概率与误差 | |
| result = model.predict_proba(image) | |
| print('属于该类的概率与误差是:', result) | |
| # 给出类别预测:0或者1 | |
| result = model.predict_classes(image) | |
| # 返回类别预测结果 | |
| return result[0] | |
| ####如果带不动就展示这个程序咯 | |
| if __name__ == '__main__': | |
| # 加载模型 | |
| my_model = keras.models.load_model('D:\model\model.h5') | |
| # 框住人脸的矩形边框颜色 | |
| color = (0, 255, 0) | |
| # 捕获指定摄像头的实时视频流 | |
| camera = cv2.VideoCapture(0) | |
| # 人脸识别分类器本地存储路径 | |
| cascade_path = "/anaconda3/share/OpenCV/haarcascades/haarcascade_frontalface_alt2.xml" | |
| # 循环检测识别人脸 | |
| while True: | |
| ret, frame = camera.read() # 读取一帧视频 | |
| # 图像灰化,降低计算复杂度 | |
| gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY) | |
| # 使用人脸识别分类器,读入分类器 | |
| face_cascade = cv2.CascadeClassifier(cv2.data.haarcascades + 'haarcascade_frontalface_default.xml') | |
| # 利用分类器识别出哪个区域为人脸 | |
| faces = face_cascade.detectMultiScale(gray, 1.1, 5) | |
| if len(faces) > 0: | |
| for (x, y, w, h) in faces: | |
| # 截取脸部图像提交给模型识别这是谁 | |
| image = frame[y: y + h, x: x + w] | |
| faceID = face_predict(image,my_model) | |
| # 如果是“我” | |
| if faceID == 0: | |
| cv2.rectangle(frame, (x, y), (x + w, y + h), color, thickness=2) | |
| # 文字提示是谁 | |
| cv2.putText(frame, 'me', | |
| (x + 30, y + 30), # 坐标 | |
| cv2.FONT_HERSHEY_SIMPLEX, # 字体 | |
| 1, # 字号 | |
| (255, 0, 255), # 颜色 | |
| 2) # 字的线宽 | |
| else: | |
| cv2.rectangle(frame, (x, y), (x + w, y + h), color, thickness=2) | |
| # 文字提示是谁 | |
| cv2.putText(frame, 'others', | |
| (x + 30, y + 30), # 坐标 | |
| cv2.FONT_HERSHEY_SIMPLEX, # 字体 | |
| 1, # 字号 | |
| (255, 0, 255), # 颜色 | |
| 2) | |
| cv2.imshow("camera", frame) | |
| # 等待1毫秒看是否有按键输入 | |
| k = cv2.waitKey(1) | |
| # 如果输入q则退出循环 | |
| if k & 0xFF == ord('q'): | |
| break | |
| # 释放摄像头并销毁所有窗口 | |
| camera.release() | |
| cv2.destroyAllWindows() |