gesturedetector.py

# import the necessary packages
from sklearn.metrics import pairwise
import numpy as np
import cv2

class GestureDetector:
    def __init__(self):
        pass

    def detect(self, thresh, cnt):
        # compute the convex hull of the contour and then find the most extreme points along
        # the hull
        hull = cv2.convexHull(cnt)
        extLeft = tuple(hull[hull[:, :, 0].argmin()][0])
        extRight = tuple(hull[hull[:, :, 0].argmax()][0])
        extTop = tuple(hull[hull[:, :, 1].argmin()][0])
        extBot = tuple(hull[hull[:, :, 1].argmax()][0])

        # compute the center (x, y)-coordinates based on the extreme points, then add a little
        # extra percentage to the y-coordinate to lower the region towards the center of the
        # palm
        cX = (extLeft[0] + extRight[0]) / 2
        cY = (extTop[1] + extBot[1]) / 2
        cY += (cY * 0.15)
        cY = int(cY)
        cX = int(cX)
        # compute the distances between the centroid and the extreme coordinates, then find the
        # largest distance, and use it to complete radius of palm region
        D = pairwise.euclidean_distances([(cX, cY)], Y=[extLeft, extRight, extTop, extBot])[0]
        maxDist = D[D.argmax()]
        r = int(0.7 * maxDist)
        circum = 2 * np.pi * r

        # construct the circular ROI that includes the palm + fingers
        circleROI = np.zeros(thresh.shape[:2], dtype="uint8")
        cv2.circle(circleROI, (cX, cY), r, 255, 1)
        circleROI = cv2.bitwise_and(thresh, thresh, mask=circleROI)

        # find contours in the circular ROI and initialize the total number of fingers counted
        # in the frame
        (_, cnts, _) = cv2.findContours(circleROI.copy(), cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_NONE)
        total = 0

        # loop over the contours
        for c in cnts:
            # compute the bounding box of the contour
            (x, y, w, h) = cv2.boundingRect(c)

            # increment the total number of fingers only if (1) the number of points along the
            # contour does not exceed 25% of the circumference and (2) the contour region is not
            # at the bottom of the circle (which is the wrist area)
            if c.shape[0] < circum * 0.25 and (y + h) < cY + (cY * 0.25):
                total += 1

        # return the total number of fingers detected
        return total

    @staticmethod
    def drawText(roi, i, val, color=(0, 0, 255)):
        # draw the text on the ROI
        cv2.putText(roi, str(val), ((i * 50) + 20, 45), cv2.FONT_HERSHEY_SIMPLEX, 1.0,
            color, 3)

    @staticmethod
    def drawBox(roi, i, color=(0, 0, 255)):
        # draw the box on the ROI
        cv2.rectangle(roi, ((i * 50) + 10, 10), ((i * 50) + 50, 60), color, 2)