V8N = 0

V11N = 1

V11S = 2

def __init__(self, model: YoloModel):

super().__init__()

self.model = self.build_model(model)

def build_model(self, model: YoloModel) -> Any:

"""

# Import the optional Ultralytics YOLO model

try:

from ultralytics import YOLO

except ModuleNotFoundError as e:

raise ImportError(

"Yolo is an optional detector, ensure the library is installed. "

"Please install using 'pip install ultralytics'"

) from e

weight_file = r'G:/ai_dotnet/UserSmartBoard/Assets/deepface' + '/' + WEIGHT_NAMES[model.value]

# Return face_detector

return YOLO(weight_file)

def export_onnx(self, out_path: str):

self.model.export(format="onnx")

def detect_faces(self, img: np.ndarray) -> List[FacialAreaRegion]:

"""

resp = []

# Detect faces

results = self.model.predict(

img,

verbose=False,

show=False,

conf=float(os.getenv("YOLO_MIN_DETECTION_CONFIDENCE", "0.25")),

)[0]

# For each face, extract the bounding box, the landmarks and confidence

for result in results:

if result.boxes is None:

continue

# Extract the bounding box and the confidence

x, y, w, h = result.boxes.xywh.tolist()[0]

confidence = result.boxes.conf.tolist()[0]

right_eye = None

left_eye = None

# yolo-facev8 is detecting eyes through keypoints,

# while for v11 keypoints are always None

if result.keypoints is not None:

# right_eye_conf = result.keypoints.conf[0][0]

# left_eye_conf = result.keypoints.conf[0][1]

right_eye = result.keypoints.xy[0][0].tolist()

left_eye = result.keypoints.xy[0][1].tolist()

# eyes are list of float, need to cast them tuple of int

left_eye = tuple(int(i) for i in left_eye)

right_eye = tuple(int(i) for i in right_eye)

x, y, w, h = int(x - w / 2), int(y - h / 2), int(w), int(h)

facial_area = FacialAreaRegion(

x=x,

y=y,

w=w,

h=h,

left_eye=left_eye,

right_eye=right_eye,

confidence=confidence,

)

resp.append(facial_area)

# 初始化一个FaceDetector对象

yolo = YoloDetectorClientV8n()

# 读取图像

img = cv2.imread(r"G:\\ai_dotnet\\UserSmartBoard\\SimpleData\\111.jpg")

# 检测人脸

faces = yolo.detect_faces(img)

# 打印检测到的人脸数量和详细信息

print(f"检测到{len(faces)}张人脸。")

# 在图像上绘制检测结果

for face in faces:

# 绘制人脸边界框

cv2.rectangle(img, (face.x, face.y), (face.x + face.w, face.y + face.h), (0, 255, 0), 2)

# 绘制左眼

cv2.circle(img, face.left_eye, 2, (255, 0, 0), 2)

# 绘制右眼

cv2.circle(img, face.right_eye, 2, (0, 0, 255), 2)

# 添加置信度文本

cv2.putText(img, f"{face.confidence:.2%}", (face.x, face.y - 10),

cv2.FONT_HERSHEY_SIMPLEX, 0.5, (0, 255, 0), 2)

# 保存结果图像

cv2.imwrite("out.jpg", img)

# 打印检测信息

for i, face in enumerate(faces, 1):

print(f"\n人脸 #{i}:")

print(f"位置: x={face.x}, y={face.y}, 宽度={face.w}, 高度={face.h}")

print(f"置信度: {face.confidence:.2%}")

print(f"左眼坐标: {face.left_eye}")

import onnxruntime as ort

import numpy as np

import cv2

from FacialAreaRegion import FacialAreaRegion

weight_file = r'G:/ai_dotnet/UserSmartBoard/Assets/deepface' + '/' + "yolov8n-face.onnx"

# Load the ONNX model

model = ort.InferenceSession(weight_file)

# Load the image

img_path = "G:\\ai_dotnet\\UserSmartBoard\\SimpleData\\111.jpg"

img = cv2.imread(img_path)

orig_img = img.copy() # 保存原始图像用于绘制

orig_h, orig_w = img.shape[:2]

img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)

# Preprocess the image

input_size = 640

img = cv2.resize(img, (input_size, input_size))

img = np.transpose(img, (2, 0, 1))

img = np.expand_dims(img, axis=0)

img = img.astype(np.float32) / 255.0

# Run the model

outputs = model.run(None, {"images": img})

# 解析模型输出 [1, 20, 8400]

predictions = outputs[0]

# 重新排列输出形状为 [8400, 20]

predictions = predictions[0].T

# 设置置信度阈值

conf_threshold = 0.5

# 创建列表存储有效的检测结果

boxes = []

confidences = []

keypoints = []

# 遍历所有预测框

for i in range(predictions.shape[0]):

# 提取置信度

confidence = predictions[i][4]

# 过滤低置信度的检测结果

if confidence >= conf_threshold:

# 提取边界框参数

x_center, y_center, width, height = predictions[i][0:4]

# 将边界框转换到原始图像尺寸

x1 = int((x_center - width/2) * orig_w / input_size)

y1 = int((y_center - height/2) * orig_h / input_size)

w = int(width * orig_w / input_size)

h = int(height * orig_h / input_size)

# 确保边界框在图像范围内

x1 = max(0, x1)

y1 = max(0, y1)

# 提取5个关键点

kpts = []

for j in range(5):

kp_x = predictions[i][5 + j*3] * orig_w / input_size

kp_y = predictions[i][6 + j*3] * orig_h / input_size

kp_conf = predictions[i][7 + j*3]

kpts.append((int(kp_x), int(kp_y), kp_conf))

boxes.append([x1, y1, w, h])

confidences.append(float(confidence))

keypoints.append(kpts)

# 应用非极大值抑制 (NMS)

indices = cv2.dnn.NMSBoxes(boxes, confidences, conf_threshold, 0.4)

# 创建面部区域对象列表

face_regions = []

# 遍历保留的检测结果

for i in indices:

box = boxes[i]

conf = confidences[i]

kpts = keypoints[i]

# 提取坐标

x, y, w, h = box

# 提取关键点坐标 (YOLOv8-face模型中0和1为左右眼)

right_eye = (kpts[0][0], kpts[0][1])

left_eye = (kpts[1][0], kpts[1][1])

# 创建面部区域对象

facial_area = FacialAreaRegion(

x=x,

y=y,

w=w,

h=h,

left_eye=left_eye,

right_eye=right_eye,

confidence=conf

)

face_regions.append(facial_area)

print(f"ONNX模型检测到{len(face_regions)}张人脸")

# 在原图上绘制结果

for face in face_regions:

# 绘制边界框

cv2.rectangle(orig_img, (face.x, face.y), (face.x + face.w, face.y + face.h), (0, 255, 0), 2)

# 绘制所有关键点

for idx, i in enumerate(indices):

kpts = keypoints[i]

# 分别绘制五个关键点(右眼、左眼、鼻子、右嘴角、左嘴角)

colors = [(0, 0, 255), (255, 0, 0), (0, 255, 0), (255, 255, 0), (0, 255, 255)]

for j, (x, y, conf) in enumerate(kpts):

if conf > 0.5: # 仅绘制置信度高的关键点

cv2.circle(orig_img, (int(x), int(y)), 3, colors[j], -1)

# 添加置信度文本

cv2.putText(orig_img, f"{face.confidence:.2f}", (face.x, face.y - 10),

cv2.FONT_HERSHEY_SIMPLEX, 0.5, (0, 255, 0), 2)

# 保存结果

cv2.imwrite("onnx_detection_result.jpg", orig_img)

# 输出检测到的人脸信息

for i, face in enumerate(face_regions):

print(f"\n人脸 #{i+1}:")

print(f"位置: x={face.x}, y={face.y}, 宽度={face.w}, 高度={face.h}")

print(f"置信度: {face.confidence:.2f}")

print(f"左眼坐标: {face.left_eye}")

print(f"右眼坐标: {face.right_eye}")