def add_entry(out, path, boxes, points, mirror):
'''add the image, points, boxes to the output file'''
w = image.shape[1]
# handle xml file
if isinstance(out, Xml):
out.append(path, boxes, [points])
if mirror:
path = utils.mirror_image(stack[0], path)
points = [(w - x, y) for x, y in points]
out.append(path, boxes, [points])
else:
out.write(f'{path}\n')
for x, y in points:
out.write(f'{x} {y}')
out.write("\n")
if mirror:
path = utils.mirror_image(stack[0], path)
out.write(f'{path}\n')
for x, y in points:
out.write(f'{w - x} {y}')
out.write("\n")
def get_random_sample(image, shape, rotation_stddev=10):
# Read a random image with landmarks and bb
image = menpo.image.Image(image.transpose((2, 0, 1)), copy=False)
image.landmarks['PTS'] = PointCloud(shape)
if np.random.rand() < .5:
image = utils.mirror_image(image)
if np.random.rand() < .5:
theta = np.random.normal(scale=rotation_stddev)
rot = menpo.transform.rotate_ccw_about_centre(
image.landmarks['PTS'], theta)
image = image.warp_to_shape(image.shape, rot)
bb = image.landmarks['PTS'].bounding_box().points
miny, minx = np.min(bb, 0)
maxy, maxx = np.max(bb, 0)
bbsize = max(maxx - minx, maxy - miny)
center = [(miny + maxy) / 2., (minx + maxx) / 2.]
shift = (np.random.rand(2) - 0.5) * 0.6 * bbsize
image.landmarks['bb'] = PointCloud([
[
center[0] - bbsize * 0.5 + shift[0],
center[1] - bbsize * 0.5 + shift[1]
],
[
center[0] + bbsize * 0.5 + shift[0],
center[1] + bbsize * 0.5 + shift[1]
],
]).bounding_box()
proportion = 1.0 / 6.0 + float(np.random.rand() - 0.5) / 6.
image = image.crop_to_landmarks_proportion(proportion, group='bb')
image = image.resize((112, 112))
random_image = image.pixels.transpose(1, 2, 0).astype('float32')
random_shape = image.landmarks['PTS'].points.astype('float32')
return random_image, random_shape
def get_mirrored_image(image, shape, init):
# Read a random image with landmarks and bb
image_m = menpo.image.Image(image.transpose((2, 0, 1)))
image_m.landmarks['init'] = PointCloud(init)
image_m = utils.mirror_image(image_m)
mirrored_image = image_m.pixels.transpose(1, 2,
0).astype('float32')
mirrored_init = image_m.landmarks['init'].points.astype('float32')
return image, init, mirrored_image, mirrored_init, shape
def load_image(path,
reference_shape,
is_training=False,
group='PTS',
mirror_image=False):
"""Load an annotated image.
In the directory of the provided image file, there
should exist a landmark file (.pts) with the same
basename as the image file.
Args:
path: a path containing an image file.
reference_shape: a numpy array [num_landmarks, 2]
is_training: whether in training mode or not.
group: landmark group containing the grounth truth landmarks.
mirror_image: flips horizontally the image's pixels and landmarks.
Returns:
pixels: a numpy array [width, height, 3].
estimate: an initial estimate a numpy array [68, 2].
gt_truth: the ground truth landmarks, a numpy array [68, 2].
"""
im = mio.import_image(path)
bb_root = im.path.parent.relative_to(im.path.parent.parent.parent)
if 'set' not in str(bb_root):
bb_root = im.path.parent.relative_to(im.path.parent.parent)
im.landmarks['bb'] = mio.import_landmark_file(
str(Path('bbs') / bb_root / (im.path.stem + '.pts')))
im = im.crop_to_landmarks_proportion(0.3, group='bb')
reference_shape = PointCloud(reference_shape)
bb = im.landmarks['bb'].lms.bounding_box()
im.landmarks['__initial'] = align_shape_with_bounding_box(
reference_shape, bb)
im = im.rescale_to_pointcloud(reference_shape, group='__initial')
if mirror_image:
im = utils.mirror_image(im)
lms = im.landmarks[group].lms
initial = im.landmarks['__initial'].lms
# if the image is greyscale then convert to rgb.
pixels = grey_to_rgb(im).pixels.transpose(1, 2, 0)
gt_truth = lms.points.astype(np.float32)
estimate = initial.points.astype(np.float32)
return pixels.astype(np.float32).copy(), gt_truth, estimate
def get_random_sample(image, shape, rotation_stddev=10):
# Read a random image with landmarks and bb
image = menpo.image.Image(image.transpose((2, 0, 1)), copy=False)
image.landmarks['PTS'] = PointCloud(shape)
if np.random.rand() < .5:
image = utils.mirror_image(image)
if np.random.rand() < .5:
theta = np.random.normal(scale=rotation_stddev)
rot = menpo.transform.rotate_ccw_about_centre(
image.landmarks['PTS'], theta)
image = image.warp_to_shape(image.shape, rot)
bb = image.landmarks['PTS'].bounding_box().points
miny, minx = np.min(bb, 0)
maxy, maxx = np.max(bb, 0)
bbsize = max(maxx - minx, maxy - miny)
center = [(miny + maxy) / 2., (minx + maxx) / 2.]
shift = (np.random.rand(2) - 0.5) / 6. * bbsize
image.landmarks['bb'] = PointCloud([
[
center[0] - bbsize * 0.5 + shift[0],
center[1] - bbsize * 0.5 + shift[1]
],
[
center[0] + bbsize * 0.5 + shift[0],
center[1] + bbsize * 0.5 + shift[1]
],
]).bounding_box()
proportion = 1.0 / 6.0 + float(np.random.rand() - 0.5) / 10.0
image = image.crop_to_landmarks_proportion(proportion, group='bb')
image = image.resize((112, 112))
random_image = image.pixels.transpose(1, 2, 0).astype('float32')
random_shape = image.landmarks['PTS'].points.astype('float32')
# Occlude
_O_AREA = 0.15
_O_MIN_H = 0.15
_O_MAX_H = 1.0
if np.random.rand() < .3:
rh = min(
112,
int((np.random.rand() * (_O_MAX_H - _O_MIN_H) + _O_MIN_H) *
112))
rw = min(112, int(12544 * _O_AREA / rh))
dy = int(np.random.rand() * (112 - rh))
dx = int(np.random.rand() * (112 - rw))
idx = int(np.random.rand() * _num_negatives)
random_image[dy:dy + rh, dx:dx + rw] = np.minimum(
1.0, _negatives[idx][dy:dy + rh, dx:dx + rw])
return random_image, random_shape
def get_random_sample(num, rotation_stddev=10):
# idx = np.random.randint(low=0, high=len(_images))
images = []
shapes = []
inits =[]
shape_3D = []
inits_3D = []
for i in range(FLAGS.batch_size):
rand_num = np.random.randint(0, num)
pixel_list = []
shape_list = []
# print('get_random_set:', idx)
im = menpo.image.Image(_images[rand_num].transpose(2, 0, 1), copy=False)
lms = _shapes[rand_num]
init = _inits[rand_num]
im.landmarks['PTS'] = lms
im.landmarks['inital'] = init
if np.random.rand() < .5:
im = utils.mirror_image(im)
if np.random.rand() < .5:
theta = np.random.normal(scale=rotation_stddev)
rot = menpo.transform.rotate_ccw_about_centre(lms, theta)
im = im.warp_to_shape(im.shape, rot)
pixels = im.pixels.transpose(1, 2, 0).astype('float32')
shape = im.landmarks['PTS'].lms.points.astype('float32')
init_2 = im.landmarks['inital'].lms.points.astype('float32')
pixel_list.append(pixels)
pixel_list = np.array(pixel_list)
inits.append(init_2)
images.append(pixel_list)
shapes.append(shape)
return images, shapes, inits
def load_image(path, reference_shape, is_training=False, group='PTS',
mirror_image=False):
"""Load an annotated image.
In the directory of the provided image file, there
should exist a landmark file (.pts) with the same
basename as the image file.
Args:
path: a path containing an image file.
reference_shape: a numpy array [num_landmarks, 2]
is_training: whether in training mode or not.
group: landmark group containing the grounth truth landmarks.
mirror_image: flips horizontally the image's pixels and landmarks.
Returns:
pixels: a numpy array [width, height, 3].
estimate: an initial estimate a numpy array [68, 2].
gt_truth: the ground truth landmarks, a numpy array [68, 2].
"""
im = mio.import_image(path)
bb_root = im.path.parent.relative_to(im.path.parent.parent.parent)
if 'set' not in str(bb_root):
bb_root = im.path.parent.relative_to(im.path.parent.parent)
im.landmarks['bb'] = mio.import_landmark_file(str(Path('bbs') / bb_root / (
im.path.stem + '.pts')))
im = im.crop_to_landmarks_proportion(0.3, group='bb')
reference_shape = PointCloud(reference_shape)
bb = im.landmarks['bb'].lms.bounding_box()
im.landmarks['__initial'] = align_shape_with_bounding_box(reference_shape,
bb)
im = im.rescale_to_pointcloud(reference_shape, group='__initial')
if mirror_image:
im = utils.mirror_image(im)
lms = im.landmarks[group].lms
initial = im.landmarks['__initial'].lms
# if the image is greyscale then convert to rgb.
pixels = grey_to_rgb(im).pixels.transpose(1, 2, 0)
gt_truth = lms.points.astype(np.float32)
estimate = initial.points.astype(np.float32)
return pixels.astype(np.float32).copy(), gt_truth, estimate
def get_random_sample(rotation_stddev=10):
idx = np.random.randint(low=0, high=len(_images))
im = menpo.image.Image(_images[idx].transpose(2, 0, 1), copy=False)
lms = _shapes[idx]
im.landmarks['PTS'] = lms
if np.random.rand() < .5:
im = utils.mirror_image(im)
if np.random.rand() < .5:
theta = np.random.normal(scale=rotation_stddev)
rot = menpo.transform.rotate_ccw_about_centre(lms, theta)
im = im.warp_to_shape(im.shape, rot)
pixels = im.pixels.transpose(1, 2, 0).astype('float32')
shape = im.landmarks['PTS'].lms.points.astype('float32')
return pixels, shape
def get_random_sample(rotation_stddev=10):
idx = np.random.randint(low=0, high=len(_images))
im = menpo.image.Image(_images[idx].transpose(2, 0, 1), copy=False)
lms = _shapes[idx]
im.landmarks['PTS'] = lms
if np.random.rand() < .5:
im = utils.mirror_image(im)
if np.random.rand() < .5:
theta = np.random.normal(scale=rotation_stddev)
rot = menpo.transform.rotate_ccw_about_centre(lms, theta)
im = im.warp_to_shape(im.shape, rot)
pixels = im.pixels.transpose(1, 2, 0).astype('float32')
shape = im.landmarks['PTS'].lms.points.astype('float32')
return pixels, shape
def process_images(queue, i, augment, paths):
print('begin p{}'.format(i), os.getpid(), os.getppid())
cnt = 0
for path in paths:
mp_image = mio.import_image(path)
landmark_path = path.parent.parent / 'Fix3' / (path.stem + '.txt')
mp_image.landmarks['PTS'] = mshape.PointCloud(np.genfromtxt(landmark_path)[:, [1, 0]])
for j in range(augment):
try:
mp_image_i = mp_image.copy()
if j % 2 == 1:
mp_image_i = utils.mirror_image(mp_image_i)
if np.random.rand() < .5:
theta = np.random.normal(scale=20)
rot = menpo.transform.rotate_ccw_about_centre(mp_image_i.landmarks['PTS'], theta)
mp_image_i = mp_image_i.warp_to_shape(mp_image_i.shape, rot, warp_landmarks=True)
# Bounding box perturbation
bb = mp_image_i.landmarks['PTS'].bounding_box().points.astype(np.float32)
miny, minx = np.min(bb, 0)
maxy, maxx = np.max(bb, 0)
bbsize = max(maxx - minx, maxy - miny)
center = [(miny + maxy) / 2., (minx + maxx) / 2.]
shift = np.random.normal(0, 0.05, 2) * bbsize
proportion = (1.0 / 6.0 + float(np.random.normal(0, 0.15))) * bbsize
mp_image_i.landmarks['bb'] = mshape.PointCloud(
[
[
center[0] - bbsize * 0.5 - proportion + shift[0],
center[1] - bbsize * 0.5 - proportion + shift[1]
],
[
center[0] + bbsize * 0.5 + proportion + shift[0],
center[1] + bbsize * 0.5 + proportion + shift[1]
],
]
).bounding_box()
# Padding, Crop, Resize
pady = int(
max(
-min(center[0] - bbsize * 0.5 - proportion + shift[0], 0),
max(center[0] + bbsize * 0.5 + proportion + shift[0] - mp_image_i.height, 0)
)
) + 100
padx = int(
max(
-min(center[1] - bbsize * 0.5 - proportion + shift[1], 0),
max(center[1] + bbsize * 0.5 + proportion + shift[1] - mp_image_i.width, 0)
)
) + 100
c, h, w = mp_image_i.pixels.shape
pad_image = np.random.rand(c, h + pady + pady, w + padx + padx)
pad_image[:, pady: pady + h, padx: padx + w] = mp_image_i.pixels.astype(np.float32)
pad_shape = mp_image_i.landmarks['PTS'].points.astype(np.float32) + np.array([pady, padx])
pad_bb = mp_image_i.landmarks['bb'].points.astype(np.float32) + np.array([pady, padx])
mp_image_i = menpo.image.Image(pad_image)
mp_image_i.landmarks['PTS'] = mshape.PointCloud(pad_shape)
mp_image_i.landmarks['bb'] = mshape.PointCloud(pad_bb).bounding_box()
mp_image_i = mp_image_i.crop_to_landmarks_proportion(0, group='bb')
mp_image_i = mp_image_i.resize((112, 112))
mp_image_i = grey_to_rgb(mp_image_i)
image = mp_image_i.pixels.transpose(1, 2, 0).astype(np.float32)
shape = mp_image_i.landmarks['PTS'].points.astype(np.float32)
except Exception as e:
traceback.print_exc()
raise e
done = False
while not done:
try:
queue.put_nowait((image, shape))
done = True
except Exception:
print('p{} wait'.format(i))
traceback.print_exc()
time.sleep(0.5)
cnt += 1
print('calc{} done {}/{}'.format(i, cnt, len(paths)))
print('end p{}'.format(i), len(paths))