def ChangeBackg2Gray(PathInput, Pathoutput):
"""
Get a video and grayscale the background leaving the objects in colors
:param PathInput: string. Path input
:param Pathoutput: string. Path output
:return: Nothing
"""
cap = cv2.VideoCapture(PathInput)
fps = int(cap.get(cv2.CAP_PROP_FPS))
change_bg = alter_bg(model_type="pb")
change_bg.load_pascalvoc_model("xception_pascalvoc.pb")
change_bg.gray_video(PathInput,
frames_per_second=fps,
output_video_name=Pathoutput,
detect='person')
def ChangeBackground(BackgroundImage, PathInput, Pathoutput):
"""
Change the background of a video to a specific image. The objects in the frames do not change
:param BackgroundImage: np.ndarray . Image to set as the background
:param PathInput: string. Path input
:param Pathoutput: string. Path output
:return: Nothing
"""
cap = cv2.VideoCapture(PathInput)
fps = int(cap.get(cv2.CAP_PROP_FPS))
change_bg = alter_bg(model_type="pb")
change_bg.load_pascalvoc_model("xception_pascalvoc.pb")
change_bg.change_video_bg(PathInput,
BackgroundImage,
frames_per_second=fps,
output_video_name=Pathoutput,
detect='person')
# lets begin.
# first , lets download the model :
# model link :
# https://github.com/ayoolaolafenwa/PixelLib/releases/download/1.1/xception_pascalvoc.pb
# copy the file to c:/models
import pixellib
from pixellib.tune_bg import alter_bg
import cv2
cam = cv2.VideoCapture(0) # connet to camera
# set the captue resolution
cam.set(3,1920)
cam.set(4,1080)
change_bg = alter_bg(model_type="pb")
change_bg.load_pascalvoc_model('c:/models/xception_pascalvoc.pb')
# press q to exit the camera
seg , result = change_bg.color_camera(cam, frames_per_second=10, colors = (200,129,160), show_frames=True, frame_name="frame", detect="person", output_video_name="c:/demo/cameraSolidColor.mp4")
# in this sample we change the background color to white : 255,255,255 and we save it to a file : "c:/demo/cameraSolidColor.mp4"
#lets change to another color
def create_xml_from_image(self, output_loc, classification_name):
"""Method to extract largest identified object in image using
FasterRCNN. Method provides both a photo of image with a bounding box
and a associate XML PASCAL VOC file for every image that contains
the bounding box information.
Args:
output_loc: Folder containing all the images
Returns:
None
"""
for filename in os.listdir(output_loc):
if filename.endswith(".jpg") or filename.endswith(".png"):
#Use FasterRCNN to find the drone and segment it to a white background
change_bg = alter_bg(model_type="pb")
change_bg.load_pascalvoc_model("xception_pascalvoc.pb")
print(output_loc + filename)
change_bg.color_bg(output_loc + '/' + filename,
colors=(255, 255, 255),
output_image_name=output_loc +
'/boxed/boxed' + filename)
#Read in segmented image
frame = cv2.imread(output_loc + '/boxed/boxed' + filename)
# Our operations on the frame come here
gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
gray = cv2.GaussianBlur(gray, (7, 7), 0)
gray = cv2.medianBlur(gray, 3) #to remove salt and paper noise
#Binary threshold
ret, thresh = cv2.threshold(gray, 200, 255, 0)
#Outer boundery only
kernel = np.ones((2, 2), np.uint8)
thresh = cv2.morphologyEx(thresh, cv2.MORPH_GRADIENT, kernel)
#to strength week pixels
thresh = cv2.dilate(thresh, kernel, iterations=5)
contours, hierarchy = cv2.findContours(thresh,
cv2.RETR_EXTERNAL,
cv2.CHAIN_APPROX_NONE)
#Draw bounding box
cv2.drawContours(frame, contours, -1, (0, 255, 0), 5)
# find the biggest countour (c) by the area
c = max(contours, key=cv2.contourArea) if contours else None
x, y, w, h = cv2.boundingRect(c)
# draw the biggest contour (c) in blue
cv2.rectangle(frame, (x, y), (x + w, y + h), (255, 0, 0), 2)
# Write resulting frame
cv2.imwrite(output_loc + '/boxed/boxed' + filename, frame)
#Variables
xmin = x
xmax = x + w
ymin = y
ymax = y + h
width, height, depth = frame.shape
#Create new XML file in PASCAL VOC format
with open('box_output.xml', encoding='latin-1') as f:
tree = ET.parse(f)
root = tree.getroot()
#Replace values in XML file
for elem in root.getiterator():
try:
elem.text = elem.text.replace(
'foldername_', output_loc)
elem.text = elem.text.replace(
'filename_', filename)
elem.text = elem.text.replace('width_', str(width))
elem.text = elem.text.replace(
'height_', str(height))
elem.text = elem.text.replace('depth_', str(depth))
elem.text = elem.text.replace(
'Drone_name', str(classification_name))
elem.text = elem.text.replace('xmin_', str(xmin))
elem.text = elem.text.replace('xmax_', str(xmax))
elem.text = elem.text.replace('ymin_', str(ymin))
elem.text = elem.text.replace('ymax_', str(ymax))
except AttributeError:
pass
#Output new XML file
tree.write(output_loc + '/' + filename[:-4] + '.xml',
encoding='latin-1')
else:
continue
#Allow user to scan through images in the 'boxed' for mistakes and have those files erased
while True:
if input(
"Please take this time to scan through 'images/boxed' folder for images where the bounding box is not correctly identifying the drone. Please delete all misclassified images. Once that is complete please press Enter"
) != 'y':
break
#Remove corresponding jpg/xml files that users deleted from the bounding box images folder
boxed_files = [str(x)[5:-3] for x in os.listdir(output_loc + '/boxed')]
print(boxed_files)
for filename in os.listdir(output_loc):
if filename.endswith(".xml"):
if filename[:-3] not in boxed_files:
try:
os.remove(output_loc + '/' + filename)
except:
continue
#Remove jpg files where bounding boxes were not present
list_files = [str(x)[:-3] for x in os.listdir(output_loc)]
single_files = [i for i in list_files if list_files.count(i) == 1]
for file in single_files:
try:
os.remove(output_loc + '/' + file + 'jpg')
except:
continue
#Break the images and XML files into train, test, and valid folders
jpg_count = 1
for filename in os.listdir(output_loc):
if filename.endswith(".jpg") or filename.endswith(".png"):
if jpg_count % 2 == 0:
shutil.move(output_loc + '/' + str(filename),
output_loc + '/train/' + str(filename))
shutil.move(
output_loc + '/' + str(filename)[:-3] + 'xml',
output_loc + '/train/' + str(filename)[:-3] + 'xml')
jpg_count += 1
elif str(jpg_count).endswith('1') or str(jpg_count).endswith(
'3') or str(jpg_count).endswith('5'):
shutil.move(output_loc + '/' + str(filename),
output_loc + '/valid/' + str(filename))
shutil.move(
output_loc + '/' + str(filename)[:-3] + 'xml',
output_loc + '/valid/' + str(filename)[:-3] + 'xml')
jpg_count += 1
else:
shutil.move(output_loc + '/' + str(filename),
output_loc + '/test/' + str(filename))
shutil.move(
output_loc + '/' + str(filename)[:-3] + 'xml',
output_loc + '/test/' + str(filename)[:-3] + 'xml')
jpg_count += 1
fgbg = cv2.createBackgroundSubtractorMOG2()
while (1):
ret, frame = cap.read()
fgmask = fgbg.apply(frame)
cv2.imshow('Input', frame)
cv2.imshow('fgmask', fgmask)
k = cv2.waitKey(60) & 0xff
if k == ord("q"):
break
cap.release()
cv2.destroyAllWindows()
change_bg = alter_bg(model_type="pb") #loading_data_model
change_bg.load_pascalvoc_model("xception_pascalvoc.pb") #loading_data_model
def bgblur():
capture = cv2.VideoCapture(0)
change_bg.blur_camera(capture,
frames_per_second=10,
extreme=True,
show_frames=True,
frame_name="frame",
output_video_name="bgblur_out.mp4")
def bgcolor():
capture = cv2.VideoCapture(0)
from pixellib.tune_bg import alter_bg
change_bg = alter_bg()
print(change_bg)
change_bg.load_pascalvoc_model(
"deeplabv3_xception_tf_dim_ordering_tf_kernels.h5")
print("-=-=-")
change_bg.change_bg_img(
f_image_path="img/sample.jpg",
b_image_path="img/background.jpg",
output_image_name="img/new_img.jpg",
)
print("opopop")
import pixellib
from pixellib.tune_bg import alter_bg
effects_obj = alter_bg(model_type='pb')
effects_obj.load_pascalvoc_model('xception_pascalvoc.pb')
effects_obj.blur_bg("bedroom.png", extreme=True, output_image_name="outty.jpg")