Write a python script to define the detecting region by clicking and dragging mouse

In object tracking, usually, we need to define a rectangular box which contains the object of interest and obtain the box information as follows.

Box: (80, 86, 263, 312)

where (80, 86) is the upper-left point of box. The 263 and 312 are the width and height, respectively.

And this work can be done by writing a simple script (named tk-detect-box) as below.

tk-detect-box:

#!/usr/local/bin/python3

import cv2
import argparse
import os


class Rect:
    def __init__(self, x, y, width, height):
        self.x = x
        self.y = x
        self.width  = width
        self.height = height

def detect_box(image, win_name="window (hit q to exit)"):
    # Return a box for detection

    box_defined = False
    box = Rect(0, 0, 0, 0)
    def define_box(event, x, y, flags, param):

        nonlocal box_defined, box
        if event == cv2.EVENT_LBUTTONDOWN:
            box_defined = False
            box.x = x
            box.y = y
            box.width  = 0
            box.height = 0

        if event == cv2.EVENT_MOUSEMOVE:
            if not box_defined:
                box.width  = x - box.x
                box.height = y - box.y

        if event == cv2.EVENT_LBUTTONUP:
            box_defined = True

    def do_nothing(event, x, y, flags, param):
        pass

    # set mouse callback
    cv2.setMouseCallback(win_name, define_box)

    while True:
        # display the frame from video capture
        clone = image.copy()
        if box.x > 0 and box.width > 0:
            # Starting and ending point of the rectangle
            p0 = (box.x, box.y)
            p1 = (box.x + box.width, box.y + box.height)
            cv2.rectangle(clone, p0, p1, (0, 255, 0), thickness=2)

        cv2.imshow(win_name, clone)

        if (cv2.waitKey(1) & 0xFF) == ord('q'): # Hit `q` to exit
            break

    # Set a mouse callback which does nothing
    cv2.setMouseCallback(win_name, do_nothing)

    return box


# Construct the argument parser and parse the arguments
ap = argparse.ArgumentParser()
ap.add_argument("-i", "--image", required=True, help="image path.")
args = vars(ap.parse_args())

# Arguments
image_path = args['image']

# Determine the width and height from the first image
image = cv2.imread(image_path)

# Create a window
win_name = 'window (hit q to exit)'
cv2.namedWindow(win_name)

# Define the detecting box
box = detect_box(image, win_name)

# Output box information
print("Box: ({}, {}, {}, {}) ".format(box.x, box.y, box.width, box.height))

cv2.destroyAllWindows()
{% endcodeblock %}

The usage of this script is like

tk-detect-box -i girl.png

It will open the image and let you define a rectangular box by clicking and dragging the mouse. After hitting q and it will show out the box information.

Comments

Write a python script to define the detecting region by clicking and dragging mouse

Published

Category

Tags