def images_to_single_fruit_centroid(fruit_number, image_path):
import PIL
image = PIL.Image.open(image_path).resize((320,240), PIL.Image.NEAREST)
# fruit = (Image(image_path, grey=True)==fruit_number)*1.0
fruit = Image(image)==fruit_number
# import pdb; pdb.set_trace()
blobs = fruit.blobs()
# plt.imshow(fruit.image)
# plt.annotate(str(fruit_number), np.array(blobs[0].centroid).reshape(2,))
# plt.show()
# assert len(blobs) == 1, "Must have only 1 fruit of the given fruit type in this photo"
return np.array(blobs[0].centroid).reshape(2,) # (2,)
def get_triangulation_params(base_dir):
# Assume everything is in base_dir/workshop_output
import os
from pathlib import Path
import ast
fruit_lst_centroids = [[], [], [], []]
fruit_lst_pose = [[], [], [], []]
base_dir = Path(base_dir)
files = os.listdir(base_dir / 'workshop_output')
assert 'images.txt' in files
image_poses = {}
with open(base_dir / 'workshop_output/images.txt') as fp:
for line in fp.readlines():
pose_dict = ast.literal_eval(line)
image_poses[pose_dict['imgfname']] = pose_dict['pose']
for file_path in image_poses.keys():
img_vals = set(
Image(base_dir / file_path, grey=True).image.reshape(-1))
for fruit_num in img_vals:
if fruit_num > 0:
try:
centroid = images_to_single_fruit_centroid(
fruit_num, base_dir / file_path)
pose = image_poses[file_path]
fruit_lst_centroids[fruit_num - 1].append(centroid)
fruit_lst_pose[fruit_num - 1].append(
np.array(pose).reshape(3, ))
except ZeroDivisionError:
pass
completed_triangulations = {}
for i in range(4):
if len(fruit_lst_centroids[i]) > 0:
centroids = np.stack(fruit_lst_centroids[i], axis=1) # (2,n)
pose = np.stack(fruit_lst_pose[i], axis=1) # (3,n)
completed_triangulations[i + 1] = {
'centroids': centroids,
'pose': pose
} # entroids (2,n), pose (3, n)
return completed_triangulations
xlabel='pixel value',
ylabel='number of pixels',
**kwargs)
elif type == 'cumulative':
plot_histogram(self.xs.flatten(),
self.cs.flatten(),
block=block,
xlabel='pixel value',
ylabel='cumulative number of pixels',
**kwargs)
elif type == 'normalized':
plot_histogram(self.xs.flatten(),
self.ns.flatten(),
block=block,
xlabel='pixel value',
ylabel='normalized cumulative number of pixels',
**kwargs)
# --------------------------------------------------------------------------#
if __name__ == '__main__':
print('ImageProcessingKernel.py')
from machinevisiontoolbox import Image
im = Image('penguins.png', grey=True)
h = im.hist()
print(h)
h.plot(block=True)
>>> from machinevisiontoolbox import Image
>>> im = Image('shark2.png')
>>> blobs = im.blobs()
>>> print(blobs)
"""
return Blob(self, **kwargs)
if __name__ == "__main__":
from machinevisiontoolbox import Image
import matplotlib.pyplot as plt
im = Image('shark2.png')
blobs = im.blobs()
print(blobs)
im.disp()
blobs.plot_labelbox(color='yellow')
blobs.plot_centroid()
plt.show()
# # read image
# from machinevisiontoolbox import Image
# im = Image(cv.imread('images/multiblobs.png', cv.IMREAD_GRAYSCALE))
# # call Blobs class
# b = Blob(image=im)
# # plot image
# plot_box(centre=(300,200), wh=(40,40), fillcolor='red', alpha=0.5)
# plot_point([(200,200), (300, 300), (400,400)], marker='r*', color='blue', text="bob {}")
# plot_labelbox("hello", color='red', textcolor='white', centre=(300,300), wh=(60,40))
# plt.show()
import numpy as np
from machinevisiontoolbox import idisp, iread, Image
im = np.zeros((100, 100, 3), 'uint8')
# im, file = iread('flowers1.png')
# draw_box(im, color=(255,0,0), centre=(50,50), wh=(20,20))
# draw_point(im, [(200,200), (300, 300), (400,400)], color='blue')
# draw_labelbox(im, "box", thickness=3, centre=(100,100), wh=(100,30), color='red', textcolor='white')
idisp(im)
x = np.random.randint(0, 100, size=(10, ))
y = np.random.randint(0, 100, size=(10, ))
plot_point((x, y), 'w+')
plt.draw()
plt.show(block=True)
im = Image('penguins.png')
h = im.hist()
else:
# isint
th0 = np.round(0.1 * np.iinfo(img.dtype).max)
if th1 is None:
th1 = 1.5 * th0
# compute gradients Ix, Iy using guassian kernel
dg = self.kdgauss(sigma)
out = []
for im in img:
Ix = np.abs(im.convolve(dg, 'same'))
Iy = np.abs(im.convolve(np.transpose(dg), 'same'))
# Ix, Iy must be 16-bit input image
Ix = np.array(Ix, dtype=np.int16)
Iy = np.array(Iy, dtype=np.int16)
out.append((cv.Canny(Ix, Iy, th0, th1, L2gradient=True)))
return self.__class__(out)
# --------------------------------------------------------------------------#
if __name__ == '__main__':
print('ImageProcessingKernel.py')
from machinevisiontoolbox import Image
print(Image.kcircle(5))
# sf0 = sf[0:3]
# sf0.u
# drawing = sf.drawSiftKeypoints(imgray)
# TODO would be nice to make a root-sift descriptor method, as it is a
# simple addition to the SIFT descriptor
# test matching
# import code
# code.interact(local=dict(globals(), **locals()))
from machinevisiontoolbox import Image
kp1 = Image('eiffel2-1.png').SIFT()
kp2 = Image('eiffel2-2.png').SIFT()
matches = kp1.match(kp2)
# im = Image('eiffel2-1.png')
# ax = im.disp()
# # sort into descending order
# ks = kp1.sort()
# print(len(kp1), len(ks))
# print(kp1[0]._descriptor)
# print(ks[0]._descriptor)
# kp1.plot(hand=True, handalpha=0.2)