def __init__(self, fps: int, width: int, height: int, codec: str,
no_background: bool, background_blur: int,
background_keep_aspect: bool, use_foreground: bool,
hologram: bool, tiling: bool, background_image: str,
foreground_image: str, foreground_mask_image: str,
webcam_path: str, v4l2loopback_path: str,
use_akvcam: bool) -> None:
self.no_background = no_background
self.use_foreground = use_foreground
self.hologram = hologram
self.tiling = tiling
self.background_blur = background_blur
self.background_keep_aspect = background_keep_aspect
self.background_image = background_image
self.foreground_image = foreground_image
self.foreground_mask_image = foreground_mask_image
self.real_cam = RealCam(webcam_path, width, height, fps, codec)
# In case the real webcam does not support the requested mode.
self.width = self.real_cam.get_frame_width()
self.height = self.real_cam.get_frame_height()
self.use_akvcam = use_akvcam
if not use_akvcam:
self.fake_cam = pyfakewebcam.FakeWebcam(v4l2loopback_path,
self.width, self.height)
else:
self.fake_cam = AkvCameraWriter(v4l2loopback_path, self.width,
self.height)
self.foreground_mask = None
self.inverted_foreground_mask = None
self.images: Dict[str, Any] = {}
self.image_lock = asyncio.Lock()
self.bodypix_model = load_model(
download_model(BodyPixModelPaths.MOBILENET_FLOAT_50_STRIDE_16))
def load_bodypix_model(self) -> BodyPixModelWrapper:
LOGGER.info('loading bodypix model: %s', self.model_path)
bodypix_model = load_model(
download_model(self.model_path),
internal_resolution=self.internal_resolution)
LOGGER.info('bodypix internal resolution: %s',
bodypix_model.internal_resolution)
return bodypix_model
def __init__(self):
super().__init__('bodypix')
self.pubMask = self.create_publisher(Image, "user/body_mask", 10)
self.colMask = self.create_publisher(Image, "user/colored_body_mask",
10)
self.subImg = self.create_subscription(Image, "camera/image",
self.imageCallback, 10)
self.bodypix_model = load_model(
download_model(BodyPixModelPaths.MOBILENET_FLOAT_100_STRIDE_16))
import cv2
import numpy as np
import tensorflow as tf
from tf_bodypix.api import download_model, load_model, BodyPixModelPaths
if len(sys.argv) > 1:
file = sys.argv[1]
else:
file = "images/persons.jpg"
img = cv2.imread(file)
print(type(img))
print(img.shape)
bodypix_model = load_model(
download_model(BodyPixModelPaths.MOBILENET_FLOAT_50_STRIDE_16))
# load image file & display it
image = tf.keras.preprocessing.image.load_img(file)
#image.show() # PIL Image
image_array = tf.keras.preprocessing.image.img_to_array(image)
print(type(image_array))
image = np.array(image_array, dtype=np.uint8)
image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)
cv2.imshow('image', image)
result = bodypix_model.predict_single(image_array)
mask = result.get_mask(threshold=0.75)
#tf.keras.preprocessing.image.save_img('out/bodypix-mask.jpg',mask)
def __init__(self,
model=BodyPixModelPaths.MOBILENET_FLOAT_50_STRIDE_16) -> None:
self.bodypix_model = load_model(download_model(model))
def __init__(self):
self.bodypix_model = load_model(
download_model(BodyPixModelPaths.MOBILENET_FLOAT_50_STRIDE_16))
self.background = None
def __init__(self):
self.scale_percent = 30
self.bodypix_model = load_model(
download_model(BodyPixModelPaths.MOBILENET_FLOAT_50_STRIDE_16))