def example_face_ortho(filename_actor, filename_obj, filename_3dmarkers=None):
D = detector_landmarks.detector_landmarks(
'..//_weights//shape_predictor_68_face_landmarks.dat',
filename_3dmarkers)
image_actor = cv2.imread(filename_actor)
R = tools_GL3D.render_GL3D(filename_obj=filename_obj,
W=image_actor.shape[1],
H=image_actor.shape[0],
is_visible=False,
projection_type='O',
scale=(1, 1, 1))
R.inverce_transform_model('Z')
L = D.get_landmarks(image_actor)
L3D = D.model_68_points
rvec, tvec, scale_factor = D.get_pose_ortho(image_actor,
L,
L3D,
R.mat_trns,
do_debug=True)
print('[ %1.2f, %1.2f, %1.2f], [%1.2f, %1.2f, %1.2f], [%1.2f,%1.2f]' %
(rvec[0], rvec[1], rvec[2], tvec[0], tvec[1], tvec[2],
scale_factor[0], scale_factor[1]))
image_3d = R.get_image_ortho(rvec, tvec, scale_factor, do_debug=True)
clr = (255 * numpy.array(R.bg_color)).astype(numpy.int)
result = tools_image.blend_avg(image_actor, image_3d, clr, weight=0)
cv2.imwrite('./images/output/face_GL_ortho.png', result)
cv2.imwrite('./images/output/face_image_3d.png', image_3d)
return
def example_03_find_homography_by_keypoints(detector='SIFT', matchtype='knn'):
folder_input = 'images/ex_keypoints/'
img1 = cv2.imread(folder_input + 'left.jpg')
img2 = cv2.imread(folder_input + 'rght.jpg')
img1_gray_rgb = tools_image.desaturate(img1)
img2_gray_rgb = tools_image.desaturate(img2)
folder_output = 'images/output/'
output_filename1 = folder_output + 'left_transformed_homography.png'
output_filename2 = folder_output + 'rght_transformed_homography.png'
output_filename = folder_output + 'blended_homography.png'
if not os.path.exists(folder_output):
os.makedirs(folder_output)
else:
tools_IO.remove_files(folder_output)
points1, des1 = tools_alg_match.get_keypoints_desc(img1, detector)
points2, des2 = tools_alg_match.get_keypoints_desc(img2, detector)
homography = tools_calibrate.get_homography_by_keypoints_desc(
points1, des1, points2, des2, matchtype)
match1, match2, distance = tools_alg_match.get_matches_from_keypoints_desc(
points1, des1, points2, des2, matchtype)
for each in match1:
img1_gray_rgb = tools_draw_numpy.draw_circle(img1_gray_rgb,
int(each[1]),
int(each[0]), 3,
[0, 0, 255])
for each in match2:
img2_gray_rgb = tools_draw_numpy.draw_circle(img2_gray_rgb,
int(each[1]),
int(each[0]), 3,
[255, 255, 0])
result_image1, result_image2 = tools_calibrate.get_stitched_images_using_homography(
img1_gray_rgb,
img2_gray_rgb,
homography,
background_color=(255, 255, 255))
result_image = tools_image.blend_avg(result_image1,
result_image2,
background_color=(255, 255, 255))
# result_image = tools_calibrate.blend_multi_band(result_image1, result_image2)
cv2.imwrite(output_filename1, result_image1)
cv2.imwrite(output_filename2, result_image2)
cv2.imwrite(output_filename, result_image)
return
def example_04_blend_white_avg():
folder_input = 'images/ex_blend/'
img1 = cv2.imread(folder_input + 'white_L.png')
img2 = cv2.imread(folder_input + 'white_R.png')
folder_output = 'images/output/'
if not os.path.exists(folder_output):
os.makedirs(folder_output)
else:
tools_IO.remove_files(folder_output)
cv2.imwrite(folder_output + 'avg.png', tools_image.blend_avg(img1, img2, (255, 255, 255)))
return
def example_face_perspective(filename_actor,
filename_obj,
filename_3dmarkers=None,
do_debug=False):
D = detector_landmarks.detector_landmarks(
'..//_weights//shape_predictor_68_face_landmarks.dat',
filename_3dmarkers)
image_actor = cv2.imread(filename_actor)
image_actor = tools_image.smart_resize(image_actor, 640, 640)
R = tools_GL3D.render_GL3D(filename_obj=filename_obj,
W=image_actor.shape[1],
H=image_actor.shape[0],
is_visible=False,
projection_type='P',
scale=(1, 1, 0.25))
L = D.get_landmarks(image_actor)
L3D = D.model_68_points
L3D[:, 2] = 0
rvec, tvec = D.get_pose_perspective(image_actor, L, L3D, R.mat_trns)
print('[ %1.2f, %1.2f, %1.2f], [%1.2f, %1.2f, %1.2f]' %
(rvec[0], rvec[1], rvec[2], tvec[0], tvec[1], tvec[2]))
image_3d = R.get_image_perspective(rvec, tvec, do_debug=do_debug)
clr = (255 * numpy.array(R.bg_color)).astype(numpy.int)
result = tools_image.blend_avg(image_actor, image_3d, clr, weight=0)
cv2.imwrite('./images/output/face_GL.png', result)
M = pyrr.matrix44.multiply(pyrr.matrix44.create_from_eulers(rvec),
pyrr.matrix44.create_from_translation(tvec))
R.mat_model, R.mat_view = tools_pr_geom.decompose_model_view(M)
result = tools_render_CV.draw_points_numpy_MVP(L3D, image_actor,
R.mat_projection,
R.mat_view, R.mat_model,
R.mat_trns)
result = D.draw_landmarks_v2(result, L)
cv2.imwrite('./images/output/face_CV_MVP.png', result)
return
def blend_by_coord(image,
coord1,
coord2,
window_size,
fill_declines=False,
background_color=(128, 128, 128)):
reproject1, reproject2, m1, m2 = tools_alg_match.reproject_matches(
image,
image,
coord1,
coord2,
fill_declines=fill_declines,
window_size=window_size,
background_color=background_color)
base1 = image.copy()
mask1 = tools_image.get_mask(reproject1, background_color)
base1[mask1 == 0] = background_color
result = tools_image.blend_avg(reproject1,
base1,
background_color=background_color)
return result
def example_project_GL_vs_CV_acuro():
marker_length = 1
aperture_x, aperture_y = 0.5, 0.5
frame = cv2.imread('./images/ex_aruco/01.jpg')
R = tools_GL3D.render_GL3D(filename_obj='./images/ex_GL/box/box.obj',
W=frame.shape[1],
H=frame.shape[0],
is_visible=False,
projection_type='P')
mat_camera = tools_pr_geom.compose_projection_mat_3x3(
frame.shape[1], frame.shape[0], aperture_x, aperture_y)
axes_image, r_vec, t_vec = tools_aruco.detect_marker_and_draw_axes(
frame, marker_length, mat_camera, numpy.zeros(4))
cv2.imwrite(
'./images/output/cube_CV.png',
tools_render_CV.draw_cube_numpy(axes_image, mat_camera, numpy.zeros(4),
r_vec.flatten(), t_vec.flatten(),
(0.5, 0.5, 0.5)))
image_3d = R.get_image_perspective(r_vec.flatten(),
t_vec.flatten(),
aperture_x,
aperture_y,
scale=(0.5, 0.5, 0.5),
do_debug=True)
clr = (255 * numpy.array(R.bg_color)).astype(numpy.int)
cv2.imwrite('./images/output/cube_GL.png',
tools_image.blend_avg(frame, image_3d, clr, weight=0))
r_vec, t_vec = r_vec.flatten(), t_vec.flatten()
print('[ %1.2f, %1.2f, %1.2f], [%1.2f, %1.2f, %1.2f], %1.2f' %
(r_vec[0], r_vec[1], r_vec[2], t_vec[0], t_vec[1], t_vec[2],
aperture_x))
return
def E2E_detect_patterns(folder_in,
filename_in_field,
folder_out,
pattern_height,
pattern_width,
max_period,
window_size=25,
step=10):
print(folder_in)
image = cv2.imread(folder_in + filename_in_field)
image_gray = tools_image.desaturate(image)
peak_factor = 0.25
cut_factor = 0.95
#pairs hitmap
if os.path.isfile(folder_out + 'hitmap.png'):
hitmap = cv2.imread(folder_out + 'hitmap.png', 0)
tools_IO.remove_files(folder_out)
cv2.imwrite(folder_out + '%d-hitmap_00.png' % (window_size),
tools_image.hitmap2d_to_jet(hitmap))
cv2.imwrite(folder_out + 'hitmap.png', hitmap)
else:
tools_IO.remove_files(folder_out)
hitmap, pattern_height, pattern_width = get_self_hits(
image,
max_period,
pattern_height,
pattern_width,
window_size,
step,
folder_debug=folder_out)
cv2.imwrite(folder_out + '%d-hitmap_00.png' % (window_size),
tools_image.hitmap2d_to_jet(hitmap))
cv2.imwrite(folder_out + 'hitmap.png', hitmap)
#candidates to patterns
coord = tools_alg_grid_templates.find_local_peaks(
hitmap, int(pattern_height * peak_factor),
int(pattern_width * peak_factor))
for i in range(0, coord.shape[0]):
image_gray = tools_draw_numpy.draw_circle(image_gray, coord[i, 0],
coord[i, 1], 5, (0, 64, 255))
cv2.imwrite(folder_out + '%d-hits_00.png' % window_size, image_gray)
patterns_candidates = tools_alg_grid_templates.coordinates_to_images(
image, coord, cut_factor * pattern_height, cut_factor * pattern_width)
hitmap_candidates = tools_alg_grid_templates.coordinates_to_images(
hitmap, coord, cut_factor * pattern_height, cut_factor * pattern_width)
#best pattern
pattern = get_best_pattern(patterns_candidates, hitmap_candidates)
cv2.imwrite(folder_out + '%d-pattern_00.png' % (window_size), pattern)
#template match with pattern
hitmap = tools_alg_match.calc_hit_field_basic(image,
pattern,
rotation_tol=10,
rotation_step=0.5)
cv2.imwrite(folder_out + '%d-hitmap_01_jet.png' % (window_size),
tools_image.hitmap2d_to_jet(hitmap))
cv2.imwrite(folder_out + '%d-hitmap_01_vrd.png' % (window_size),
tools_image.hitmap2d_to_viridis(hitmap))
#visualize hits
min_value = 200
coord, image_miss = tools_alg_grid_templates.find_local_peaks_greedy(
hitmap, int(1.0 * pattern_height), int(1.0 * pattern_width), min_value)
image_gray = tools_image.desaturate(image)
for i in range(0, coord.shape[0]):
value = int(85 * (hitmap[coord[i, 0], coord[i, 1]] - min_value) /
(255 - min_value))
image_gray = tools_draw_numpy.draw_circle(
image_gray, coord[i, 0], coord[i, 1], 9,
tools_image.hsv2bgr((value, 255, 255)))
image_gray_and_miss = tools_image.put_layer_on_image(image_gray,
image_miss,
background_color=(0,
0,
0))
image_gray_and_miss = tools_image.blend_avg(image_gray,
image_gray_and_miss,
weight=0.5)
cv2.imwrite(folder_out + '%d-hits_01.png' % window_size,
image_gray_and_miss)
#reproject pattern
image_gen = generate_input(image, coord, pattern)
cv2.imwrite(folder_out + '%d-gen.png' % window_size, image_gen)
return
# ----------------------------------------------------------------------------------------------------------------------