handTrackingCamera.py

import mediapipe as mp
from google.protobuf.json_format import MessageToDict

from hand import Hand
import cv2
import numpy as np

import time 

class HandTrackingCamera():
    def __init__(self, resultQueue, cameraId, handNum):
        self.resultQueue = resultQueue

        self.running = True
        self.cameraId = cameraId
        self.handNum = handNum
        rightHand = Hand("Right")
        leftHand = Hand("Left")
        self.hands = {"Left": leftHand, "Right": rightHand}
        self.mp_hands = mp.solutions.hands
        self.mp_drawing = mp.solutions.drawing_utils
        self.mp_drawing_styles = mp.solutions.drawing_styles

        self.isHandsUpdated = {"Right": True, "Left": True}

        pass


    def updateHands(self):
        image = None

        handTrackingHandler = mp.solutions.hands.Hands(
            max_num_hands=self.handNum,
            model_complexity=0,
            min_detection_confidence=0.5,
            min_tracking_confidence=0.5)

        camera = cv2.VideoCapture(self.cameraId)

        while self.running:
            try:
                success, image = camera.read()
                if not success:
                    print("No Camera")
                else:
                    self.isHandsUpdated = {"Right": False, "Left": False}
                    image.flags.writeable = False
                    image = cv2.flip(image, 1)
                    image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)
                    image_height, image_width, _ = image.shape

                    results = handTrackingHandler.process(image)
                    
                    image.flags.writeable = True
                    image = cv2.cvtColor(image, cv2.COLOR_RGB2BGR)

                    for handInfo, handLandmarks in zip(results.multi_handedness, results.multi_hand_landmarks):
                        handName = MessageToDict(handInfo)["classification"][0]["label"]
                        # self.hands[handName].updateHandJoints(handLandmarks, image_height, image_width)
                        self.hands[handName].updateHandJoints(handLandmarks, image_height, image_width)
                        self.isHandsUpdated[handName] = True
                        
                        self.mp_drawing.draw_landmarks(
                            image,
                            handLandmarks,
                            self.mp_hands.HAND_CONNECTIONS,
                            self.mp_drawing_styles.get_default_hand_landmarks_style(),
                            self.mp_drawing_styles.get_default_hand_connections_style())

                    cv2.imshow(str(self.cameraId), image)
                    if cv2.waitKey(5) & 0xFF == 27:
                        break

                    self.resultQueue.put({"Right" : self.getRightHandJoints(), "Left": self.getLeftHandJoints()})

            except:
                pass
        

        return image

    def isHandsOverlapped(self):
        leftHandpinkyEdge = self.hands["Left"].getPinkyEdge() # ([x, y, z], [x, y, z])
        rightHandIndexEdge = self.hands["Right"].getIndexEdge()
        rightHandUpperEdge = self.hands["Right"].getUpperEdge()
        rightHandpinkyEdge = self.hands["Right"].getPinkyEdge()


        if self.isTwoLineOverlapped(leftHandpinkyEdge, rightHandIndexEdge):
            return True
        if self.isTwoLineOverlapped(leftHandpinkyEdge, rightHandUpperEdge):
            return True
        if self.isTwoLineOverlapped(leftHandpinkyEdge, rightHandpinkyEdge):
            return True 
        return False


    def isTwoLineOverlapped(self, line1, line2):
        p1, p2 = line1
        p3, p4 = line2

        if self.counterClockWise(p1, p2, p3) * self.counterClockWise(p1, p2, p4) < 0 and self.counterClockWise(p3, p4, p1) * self.counterClockWise(p3, p4, p2) < 0:
            return True
        else:
            return False


    def counterClockWise(self, p1, p2, p3):
        v1 = np.array(p2) - np.array(p1)
        v2 = np.array(p3) - np.array(p1)

        if np.cross(v1, v2) > 0:
            return 1
        else:
            return -1

    def getRightHandJoints(self):
        if self.isHandsUpdated["Right"]:
            return self.hands["Right"].getJoints()
        else:
            return None

    def getLeftHandJoints(self):
        if self.isHandsUpdated["Left"]:
            return self.hands["Left"].getJoints()
        else:
            return None

    def getRightHandJointsDiff(self):
        if self.isHandsUpdated["Right"]:
            return self.hands["Right"].calculateHandDiff()
        else:
            return None 

    def getLeftHandJointsDiff(self):
        if self.isHandsUpdated["Left"]:
            return self.hands["Left"].calculateHandDiff()
        else:
            return None


if __name__ == "__main__":
    mp_hands = mp.solutions.hands
    upperCameraHandler = mp_hands.Hands(
        max_num_hands=2,
        model_complexity=0,
        min_detection_confidence=0.5,
        min_tracking_confidence=0.5)  

    upperCamera = HandTrackingCamera(3, upperCameraHandler)

    for i in range(10000):
        upperCamera.updateHands()

        if cv2.waitKey(5) & 0xFF == 27:
            break