def __init__(self,
background_subtractor=BackgroundSubtractor(),
segmenter: MRFVideo = MRFVideo(),
descriptor: Descriptor = HistogramOfGradients()):
super().__init__(descriptor, segmenter)
self.background_subtractor = background_subtractor
self.segmenter = segmenter
def __init__(self,
descriptor: Descriptor = HistogramOfGradients(),
img_size=(30, 30),
segmenter=GrayThresh()):
PreProcessor.__init__(self)
self.segmenter = segmenter
self.descriptor = descriptor
self.roi_size = img_size
def get_codebook(
self,
images,
k_words=128,
):
descriptors = np.vstack([
HistogramOfGradients(self.winsize, self.n_bins).get_descr(img)
for img in images
])
return KMeans(n_clusters=k_words).fit(descriptors).cluster_centers_
from preprocessing.PreProcessorVideo import PreProcessorVideo
from preprocessing.representation.HistogramOfGradients import HistogramOfGradients
from preprocessing.segmentation.BackgroundSubtractor import BackgroundSubtractor
from preprocessing.segmentation.MRFVideo import MRFVideo
camera = cv2.VideoCapture(0)
# keep looping, until interrupted
keypress = cv2.waitKey(1) & 0xFF
calibrate_print_flag = calibrated_print_flag = False
roi_left = 0
roi_right = 0
roi_top = 0
roi_bottom = 0
preprocessor = PreProcessorVideo(BackgroundSubtractor(0.5), segmenter=MRFVideo(), descriptor=HistogramOfGradients())
estimator = EstimatorVideo("../../resource/models/model_hog_asl.pkl")
num_frames = 0
while keypress != ord("q"):
# observe the keypress by the user
keypress = cv2.waitKey(10) & 0xFF
# get the current frame
(grabbed, frame) = camera.read()
# resize the frame
frame = imutils.resize(frame, width=800)
# flip the frame so that it is not the mirror view
frame = cv2.flip(frame, 1)
# clone the frame
clone = frame.copy()
estimator = EstimatorVideo(
model_path='../../../resource/models/model_hog_asl.pkl')
predictions = []
score = 0
for letter_i, letter in enumerate(letters):
for i in range(1, 5):
diff_image = FileProvider.read_img(test_image_path + letter +
'/Difference' + str(i) + ".jpg")
test_image = FileProvider.read_img(test_image_path + letter +
'/Original' + str(i) + ".jpg")
if i == 1:
preprocessor = PreProcessorVideo(
background_subtractor=BackgroundSubtractorDummy(
diff_image, test_image),
descriptor=HistogramOfGradients(n_bins=16))
preprocessor.calibrate_object(test_image)
test_image = preprocessor.preprocess(test_image)
descr = preprocessor.get_descr(test_image)
estimator.stack_descr(descr)
class_ = estimator.predict()
predictions.append(class_ - 1)
if letter is letters[class_ - 1]:
score += 1
accuracy = score / 24
print("Accuracy: " + str(accuracy))
def get_descr(self, img):
return self.get_codebook_dist(
HistogramOfGradients(self.winsize, self.n_bins).get_descr(img),
self.codebook)
def __init__(self, descriptor: Descriptor = HistogramOfGradients(), segmenter: Segmenter = MRFAsl(),
img_size=(60, 60)):
self.img_size = img_size
self.descriptor = descriptor
self.segmenter = segmenter
from app.FrameHandler import FrameHandler
from app.UserInterface import UserInterface
from classification.EstimatorVideo import EstimatorVideo
from preprocessing.PreProcessorVideo import PreProcessorVideo
from preprocessing.representation.HistogramOfGradients import HistogramOfGradients
from preprocessing.segmentation.BackgroundSubtractor import BackgroundSubtractor
from preprocessing.segmentation.MRFVideo import MRFVideo
preprocessor = PreProcessorVideo(BackgroundSubtractor(0.5),
segmenter=MRFVideo(),
descriptor=HistogramOfGradients(window_size=6,
n_bins=16))
frame_handler = FrameHandler(
preprocessor,
EstimatorVideo(model_path="../../resource/examples/model_hog_asl.pkl"))
ui = UserInterface(frame_handler)
frame_handler.daemon = True
frame_handler.start()
ui.run()
import random
import cv2
from datagen.FileProviderAsl import FileProviderAsl
from preprocessing.PreProcessorAsl import PreProcessorAsl
from preprocessing.representation.HistogramOfGradients import HistogramOfGradients
from preprocessing.segmentation.MRFAsl import MRFAsl
letters = ["a", "b", "c", "d", "e", "f", "g", "h", "i", "k", "l", "m", "n", "o", "p", "q", "r", "s", "t",
"u",
"v", "w", "x", "y"]
example_image_file = "../../resource/dataset/fingerspelling5/dataset5/A/a/color_0_0027.png"
pp = PreProcessorAsl(segmenter=MRFAsl(), descriptor=HistogramOfGradients(), img_size=(120, 120))
# read image
img = FileProviderAsl.read_img(example_image_file)
cv2.imshow('depth', cv2.normalize(img[1], None, 0, 255, cv2.NORM_MINMAX))
cv2.imshow('image', img[0])
img = pp.preprocess(img)
cv2.imshow("after prefiltering", img)
descriptor = pp.get_descr(img)
print("Descriptor: \n" + str(descriptor))
# print("dim: \n" + str(descriptor.shape[1]))
cv2.waitKey(30000)
cv2.destroyAllWindows()
exit(0)