def _augment(self, img, _):
img = np.copy(img)
orig_ann = img[..., 0] # instance ID map
fixed_ann = self._fix_mirror_padding(orig_ann)
# re-cropping with fixed instance id map
crop_ann = cropping_center(fixed_ann, self.crop_shape)
orig_dst = np.zeros(orig_ann.shape, dtype=np.float32)
inst_list = list(np.unique(crop_ann))
inst_list.remove(0) # 0 is background
for inst_id in inst_list:
inst_map = np.array(fixed_ann == inst_id, np.uint8)
inst_box = bounding_box(inst_map)
# expand the box by 2px
inst_box[0] -= 2
inst_box[2] -= 2
inst_box[1] += 2
inst_box[3] += 2
inst_map = inst_map[inst_box[0]:inst_box[1],
inst_box[2]:inst_box[3]]
if inst_map.shape[0] < 2 or \
inst_map.shape[1] < 2:
continue
# chessboard distance map generation
# normalize distance to 0-1
inst_dst = distance_transform_cdt(inst_map)
inst_dst = inst_dst.astype('float32')
if self.inst_norm:
max_value = np.amax(inst_dst)
if max_value <= 0:
continue # HACK: temporay patch for divide 0 i.e no nuclei (how?)
inst_dst = (inst_dst / np.amax(inst_dst))
####
dst_map_box = orig_dst[inst_box[0]:inst_box[1],
inst_box[2]:inst_box[3]]
dst_map_box[inst_map > 0] = inst_dst[inst_map > 0]
#
img = img.astype('float32')
img = np.dstack([img, orig_dst])
return img
def _augment(self, img, _):
img = np.copy(img)
orig_ann = img[..., 0] # instance ID map
fixed_ann = self._fix_mirror_padding(orig_ann)
# re-cropping with fixed instance id map
crop_ann = cropping_center(fixed_ann, self.crop_shape)
# TODO: deal with 1 label warning
crop_ann = morph.remove_small_objects(crop_ann, min_size=30)
x_map = np.zeros(orig_ann.shape[:2], dtype=np.float32)
y_map = np.zeros(orig_ann.shape[:2], dtype=np.float32)
inst_list = list(np.unique(crop_ann))
inst_list.remove(0) # 0 is background
for inst_id in inst_list:
inst_map = np.array(fixed_ann == inst_id, np.uint8)
inst_box = bounding_box(inst_map)
# expand the box by 2px
# Because we first pad the ann at line 207, the bboxes
# will remain valid after expansion
inst_box[0] -= 2
inst_box[2] -= 2
inst_box[1] += 2
inst_box[3] += 2
inst_map = inst_map[inst_box[0]:inst_box[1],
inst_box[2]:inst_box[3]]
if inst_map.shape[0] < 2 or \
inst_map.shape[1] < 2:
continue
# instance center of mass, rounded to nearest pixel
inst_com = list(measurements.center_of_mass(inst_map))
inst_com[0] = int(inst_com[0] + 0.5)
inst_com[1] = int(inst_com[1] + 0.5)
inst_x_range = np.arange(1, inst_map.shape[1] + 1)
inst_y_range = np.arange(1, inst_map.shape[0] + 1)
# shifting center of pixels grid to instance center of mass
inst_x_range -= inst_com[1]
inst_y_range -= inst_com[0]
inst_x, inst_y = np.meshgrid(inst_x_range, inst_y_range)
# remove coord outside of instance
inst_x[inst_map == 0] = 0
inst_y[inst_map == 0] = 0
inst_x = inst_x.astype('float32')
inst_y = inst_y.astype('float32')
# normalize min into -1 scale
if np.min(inst_x) < 0:
inst_x[inst_x < 0] /= (-np.amin(inst_x[inst_x < 0]))
if np.min(inst_y) < 0:
inst_y[inst_y < 0] /= (-np.amin(inst_y[inst_y < 0]))
# normalize max into +1 scale
if np.max(inst_x) > 0:
inst_x[inst_x > 0] /= (np.amax(inst_x[inst_x > 0]))
if np.max(inst_y) > 0:
inst_y[inst_y > 0] /= (np.amax(inst_y[inst_y > 0]))
####
x_map_box = x_map[inst_box[0]:inst_box[1], inst_box[2]:inst_box[3]]
x_map_box[inst_map > 0] = inst_x[inst_map > 0]
y_map_box = y_map[inst_box[0]:inst_box[1], inst_box[2]:inst_box[3]]
y_map_box[inst_map > 0] = inst_y[inst_map > 0]
img = img.astype('float32')
img = np.dstack([img, x_map, y_map])
return img
interpolation=cv2.INTER_NEAREST)
instance_map = np.float32(instance_map)
type_map = np.float32(type_map)
#cv2.imwrite('{}/{}_inst.png'.format(save_dir, 'sample'), instance_map)
#cv2.imwrite('{}/{}_type.png'.format(save_dir, 'sample'), type_map * 60)
instance_map = cv2.resize(instance_map, (px, py),
interpolation=cv2.INTER_NEAREST)
type_map = cv2.resize(type_map, (px, py),
interpolation=cv2.INTER_NEAREST)
instance_map = np.int32(instance_map)
type_map = np.int32(type_map)
instances, instcounts = np.unique(instance_map, return_counts=True)
for i in range(1, len(instances)):
inst_size[instances[i] - 1] = instcounts[i]
inst_map = np.array(instance_map == instances[i], np.uint8)
y1, y2, x1, x2 = bounding_box(inst_map)
y1 = y1 - 2 if y1 - 2 >= 0 else y1
x1 = x1 - 2 if x1 - 2 >= 0 else x1
x2 = x2 + 2 if x2 + 2 <= inst_map.shape[1] - 1 else x2
y2 = y2 + 2 if y2 + 2 <= inst_map.shape[0] - 1 else y2
inst_map = inst_map[y1:y2, x1:x2]
contours = cv2.findContours(inst_map, cv2.RETR_TREE,
cv2.CHAIN_APPROX_SIMPLE)
#inst_size[instances[i]-1] = cv2.contourArea(contours[0])
inst_length[instances[i] - 1] = cv2.arcLength(contours[0][0], True)
indexer = np.logical_and(inst_type2 != 0, inst_size != 0)
inst_sizes = np.append(inst_sizes, inst_size[indexer])
inst_lengths = np.append(inst_lengths, inst_length[indexer])
inst_types = np.append(inst_types, inst_type2[indexer])
basename = '{}_{}_{}_{}_{}'.format(