blurring_nonhumans in video

gistfile1.txt

# import the necessary packages
from __future__ import print_function
from imutils.video import VideoStream
from imutils.object_detection import non_max_suppression
from imutils import paths
import numpy as np
import argparse
import imutils
import cv2

# INPUT
# take in the video
vc = cv2.VideoCapture('clip_personwalking.mp4')
# default start with true
success = True

# OUTPUT
# construct the argument parse and parse the arguments
ap = argparse.ArgumentParser()
ap.add_argument("-o", "--output", required=True,
	help="path to output video file")
ap.add_argument("-f", "--fps", type=int, default=20,
	help="FPS of output video")
ap.add_argument("-c", "--codec", type=str, default="avc1",
	help="codec of output video")
args = vars(ap.parse_args())

# initialize the FourCC, video writer, dimensions of the frame, and
fourcc = cv2.VideoWriter_fourcc('a', 'v', 'c', '1')
writer = None
(h, w) = (None, None)

# initialize the HOG descriptor/person detector
hog = cv2.HOGDescriptor()
hog.setSVMDetector(cv2.HOGDescriptor_getDefaultPeopleDetector())
while success:
	success, image = vc.read()

	if success:

		image = imutils.resize(image, width=1920)
		
		# copy a blurred version of the image to a new image
		blurred = cv2.medianBlur(image, 7).copy()

		if writer is None:
			(h, w) = image.shape[:2]
			writer = cv2.VideoWriter(args["output"], fourcc, args["fps"],
				(w, h), True)

		# detect people in the image
		(rects, weights) = hog.detectMultiScale(image, winStride=(4, 4),
			padding=(8, 8), scale=1.05)

		# apply non-maxima suppression to the bounding boxes using a
		# fairly large overlap threshold to try to maintain overlapping
		# boxes that are still people
		rects = np.array([[x, y, x + w, y + h] for (x, y, w, h) in rects])
		pick = non_max_suppression(rects, probs=None, overlapThresh=0.65)

		# our hacky way of masking blurry image with hi res objects
		# identify important the coordinates of objects we will avoid blurring
		# grab the the hi-fi verion and place it in the same spot as the blurred version
		for (xA, yA, xB, yB) in pick:
			cv2.rectangle(blurred, (xA, yA), (xB, yB), (0, 255, 0), 2)
			blurred[yA:yB, xA:xB] = image[yA:yB, xA:xB]

		# construct the final output image
		h=1080
		w=1920
		output = np.zeros((h, w, 3), dtype="uint8")
		output[0:h, 0:w] = blurred

		# write the output image to file
		writer.write(output)

		# if the `q` key was pressed, break from the loop
		key = cv2.waitKey(1) & 0xFF
		if key == ord("q"):
			break

writer.release()