MediaPipe简介

发表于 更新于 分类于 阅读次数：

On-device machine learning for everyone

手势识别

  • HandTrackingModule.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
import cv2
import mediapipe as mp
import time
import math

class handDetector():
    def __init__(self, mode=False, maxHands=2, detectionCon=1, trackCon=0.8):
        self.mode = mode
        self.maxHands = maxHands
        self.detectionCon = detectionCon
        self.trackCon = trackCon

        self.mpHands = mp.solutions.hands
        self.hands = self.mpHands.Hands(self.mode, self.maxHands, self.detectionCon, self.trackCon)
        self.mpDraw = mp.solutions.drawing_utils
        self.tipIds = [4, 8, 12, 16, 20]

    def findHands(self, img, draw=True):
        imgRGB = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
        self.results = self.hands.process(imgRGB)

        print(self.results.multi_handedness)  # 获取检测结果中的左右手标签并打印

        if self.results.multi_hand_landmarks:
            for handLms in self.results.multi_hand_landmarks:
                if draw:
                    self.mpDraw.draw_landmarks(img, handLms, self.mpHands.HAND_CONNECTIONS)
        return img

    def findPosition(self, img, draw=True):
        self.lmList = []
        if self.results.multi_hand_landmarks:
            for handLms in self.results.multi_hand_landmarks:
                for id, lm in enumerate(handLms.landmark):
                    h, w, c = img.shape
                    cx, cy = int(lm.x * w), int(lm.y * h)
                    # print(id, cx, cy)
                    self.lmList.append([id, cx, cy])
                    if draw:
                        cv2.circle(img, (cx, cy), 12, (255, 0, 255), cv2.FILLED)
        return self.lmList

    def fingersUp(self):
        fingers = []
        # 大拇指
        if self.lmList[self.tipIds[0]][1] > self.lmList[self.tipIds[0] - 1][1]:
            fingers.append(1)
        else:
            fingers.append(0)

        # 其余手指
        for id in range(1, 5):
            if self.lmList[self.tipIds[id]][2] < self.lmList[self.tipIds[id] - 2][2]:
                fingers.append(1)
            else:
                fingers.append(0)

        # totalFingers = fingers.count(1)
        return fingers

    def findDistance(self, p1, p2, img, draw=True, r=15, t=3):
        x1, y1 = self.lmList[p1][1:]
        x2, y2 = self.lmList[p2][1:]
        cx, cy = (x1 + x2) // 2, (y1 + y2) // 2

        if draw:
            cv2.line(img, (x1, y1), (x2, y2), (255, 0, 255), t)
            cv2.circle(img, (x1, y1), r, (255, 0, 255), cv2.FILLED)
            cv2.circle(img, (x2, y2), r, (255, 0, 255), cv2.FILLED)
            cv2.circle(img, (cx, cy), r, (0, 0, 255), cv2.FILLED)
            length = math.hypot(x2 - x1, y2 - y1)

        return length, img, [x1, y1, x2, y2, cx, cy]


def main():
    pTime = 0
    cTime = 0
    cap = cv2.VideoCapture(0)
    detector = handDetector()
    while True:
        success, img = cap.read()
        img = detector.findHands(img)        # 检测手势并画上骨架信息

        lmList = detector.findPosition(img)  # 获取得到坐标点的列表
        if len(lmList) != 0:
            print(lmList[4])

        cTime = time.time()
        fps = 1 / (cTime - pTime)
        pTime = cTime

        cv2.putText(img, 'fps:' + str(int(fps)), (10, 70), cv2.FONT_HERSHEY_PLAIN, 3, (255, 0, 255), 3)
        cv2.imshow('Image', img)
        cv2.waitKey(1)


if __name__ == "__main__":
    main()
  • AiVirtualMouse.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
import cv2
import HandTrackingModule as htm
import autopy
import numpy as np
import time

##############################
wCam, hCam = 1280, 720
frameR = 100
smoothening = 5
##############################
cap = cv2.VideoCapture(0)  # 若使用笔记本自带摄像头则编号为0  若使用外接摄像头 则更改为1或其他编号
cap.set(3, wCam)
cap.set(4, hCam)
pTime = 0
plocX, plocY = 0, 0
clocX, clocY = 0, 0

detector = htm.handDetector()
wScr, hScr = autopy.screen.size()
# print(wScr, hScr)

while True:
    success, img = cap.read()
    # 1. 检测手部 得到手指关键点坐标
    img = detector.findHands(img)
    cv2.rectangle(img, (frameR, frameR), (wCam - frameR, hCam - frameR), (0, 255, 0), 2,  cv2.FONT_HERSHEY_PLAIN)
    lmList = detector.findPosition(img, draw=False)

    # 2. 判断食指和中指是否伸出
    if len(lmList) != 0:
        x1, y1 = lmList[8][1:]
        x2, y2 = lmList[12][1:]
        fingers = detector.fingersUp()

        # 3. 若只有食指伸出 则进入移动模式
        if fingers[1] and fingers[2] == False:
            # 4. 坐标转换： 将食指在窗口坐标转换为鼠标在桌面的坐标
            # 鼠标坐标
            x3 = np.interp(x1, (frameR, wCam - frameR), (0, wScr))
            y3 = np.interp(y1, (frameR, hCam - frameR), (0, hScr))

            # smoothening values
            clocX = plocX + (x3 - plocX) / smoothening
            clocY = plocY + (y3 - plocY) / smoothening

            autopy.mouse.move(wScr - clocX, clocY)
            cv2.circle(img, (x1, y1), 15, (255, 0, 255), cv2.FILLED)
            plocX, plocY = clocX, clocY

        # 5. 若是食指和中指都伸出 则检测指头距离 距离够短则对应鼠标点击
        if fingers[1] and fingers[2]:
            length, img, pointInfo = detector.findDistance(8, 12, img)
            if length < 40:
                cv2.circle(img, (pointInfo[4], pointInfo[5]),
                           15, (0, 255, 0), cv2.FILLED)
                autopy.mouse.click()

    cTime = time.time()
    fps = 1 / (cTime - pTime)
    pTime = cTime
    cv2.putText(img, f'fps:{int(fps)}', [15, 25],
                cv2.FONT_HERSHEY_PLAIN, 2, (255, 0, 255), 2)
    cv2.imshow("Image", img)
    cv2.waitKey(1)
1
2
3
4
5
6
7
8
9
mkdir -p ~/miniconda3
curl https://repo.anaconda.com/miniconda/Miniconda3-latest-MacOSX-arm64.sh -o ~/miniconda3/miniconda.sh
bash ~/miniconda3/miniconda.sh -b -u -p ~/miniconda3
rm -rf ~/miniconda3/miniconda.sh
~/miniconda3/bin/conda init zsh

CONDA_SUBDIR=osx-64 conda create -n py37 python=3.7
conda activate py37
conda config --env --set subdir osx-64
1
2
3
pip install autopy mediapipe numpy opencv-python

python AiVirtualMouse.py

目标检测

Python

  • ObjectDetector.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
import cv2
import numpy as np
import mediapipe as mp
from mediapipe.tasks import python
from mediapipe.tasks.python import vision

MARGIN = 10  # pixels
ROW_SIZE = 10  # pixels
FONT_SIZE = 1
FONT_THICKNESS = 1
TEXT_COLOR = (255, 0, 0)  # red

IMAGE_FILE = 'image.jpg'

def visualize(
    image,
    detection_result
) -> np.ndarray:
  """Draws bounding boxes on the input image and return it.
  Args:
    image: The input RGB image.
    detection_result: The list of all "Detection" entities to be visualize.
  Returns:
    Image with bounding boxes.
  """
  for detection in detection_result.detections:
    # Draw bounding_box
    bbox = detection.bounding_box
    start_point = bbox.origin_x, bbox.origin_y
    end_point = bbox.origin_x + bbox.width, bbox.origin_y + bbox.height
    cv2.rectangle(image, start_point, end_point, TEXT_COLOR, 3)

    # Draw label and score
    category = detection.categories[0]
    category_name = category.category_name
    probability = round(category.score, 2)
    result_text = category_name + ' (' + str(probability) + ')'
    text_location = (MARGIN + bbox.origin_x,
                     MARGIN + ROW_SIZE + bbox.origin_y)
    cv2.putText(image, result_text, text_location, cv2.FONT_HERSHEY_PLAIN,
                FONT_SIZE, TEXT_COLOR, FONT_THICKNESS)

  return image

# STEP 1: Create an ObjectDetector object.
base_options = python.BaseOptions(model_asset_path='efficientdet.tflite')
options = vision.ObjectDetectorOptions(base_options=base_options,
                                       score_threshold=0.5)
detector = vision.ObjectDetector.create_from_options(options)

# STEP 2: Load the input image.
image = mp.Image.create_from_file(IMAGE_FILE)

# STEP 3: Detect objects in the input image.
detection_result = detector.detect(image)

# STEP 4: Process the detection result. In this case, visualize it.
image_copy = np.copy(image.numpy_view())
annotated_image = visualize(image_copy, detection_result)
rgb_annotated_image = cv2.cvtColor(annotated_image, cv2.COLOR_BGR2RGB)
cv2.imwrite("result.png", rgb_annotated_image)
1
2
3
4
5
wget -q -O efficientdet.tflite -q https://storage.googleapis.com/mediapipe-models/object_detector/efficientdet_lite0/int8/1/efficientdet_lite0.tflite
wget -q -O image.jpg https://storage.googleapis.com/mediapipe-tasks/object_detector/cat_and_dog.jpg

conda create -n py38 python=3.8
conda activate py38
1
2
3
pip install mediapipe==0.10.9 numpy opencv-python

python ObjectDetector.py

Android

1
2
git clone https://github.com/googlesamples/mediapipe.git
# 使用Android Studio打开项目mediapipe/examples/object_detection/android

mediapipe-01.jpg

参考