#ann = sio.loadmat(os.path.join(ann_dir, '{}.mat'.format(basename)))
#ann_inst = ann['inst_map']
#ann_type = ann['type_map']
# merge classes for CoNSeP (in paper we only utilise 3 nuclei classes and background)
# If own dataset is used, then the below may need to be modified
# TODO: move to preproc CoNSeP dataset
# ann_type[(ann_type == 3) | (ann_type == 4)] = 3
# ann_type[(ann_type == 5) | (ann_type == 6) | (ann_type == 7)] = 4
# print (f"nr_types = {cfg.nr_types}, max in annotation = {np.max(ann_type)}")
assert np.max(ann_type) <= cfg.nr_types, \
("Only {} types of nuclei are defined for training"\
"but there are {} types found in the input image.").format(cfg.nr_types, np.max(ann_type))
ann = np.dstack([ann_inst, ann_type])
ann = ann.astype('int32')
else:
# assumes that ann is HxW
# ann_inst = sio.loadmat(os.path.join(ann_dir, '{}.mat'.format(basename)))
ann = np.load(os.path.join(ann_dir, '{}.npy'.format(basename)))
ann_inst = (ann_inst.item().get('inst_map')).astype('int32')
# ann_inst = ann_inst.astype('int32')
ann = np.expand_dims(ann_inst, -1)
img = np.concatenate([img, ann], axis=-1)
sub_patches = xtractor.extract(img, cfg.extract_type)
for idx, patch in enumerate(sub_patches):
np.save("{0}/{1}_{2:03d}.npy".format(out_dir, basename, idx), patch)
print (f"{out_dir}/{basename} saved.")
if not os.path.exists(save_dir): os.mkdir(save_dir)
colormap = [class_colors[x] for x in inst_type]
overlaid_output = visualize_instances(instance_map, patchcorr,
colormap)
cv2.imwrite("{}/{}.png".format(save_dir, basename), overlaid_output)
#type_map[np.logical_and(instance_map > 0, type_map == 0)] = -1
unannot = np.logical_and(instance_map != 0, type_map == 0)
patchjoin[unannot] = 255
instance_map[unannot] = 0
ann = np.dstack([instance_map, type_map])
patchjoin = np.array(patchjoin, np.int32)
print(ann.shape)
print(patchjoin.shape)
patchjoin = np.concatenate([patchjoin, ann], axis=-1)
sub_patches = xtractor.extract(
patchjoin, extract_type
) # This extracts mirrored subpatch of 540x540 from 1k patch
for idx, patch in enumerate(sub_patches):
np.save("{}/{}_{}.npy".format(save_dir, basename, idx), patch)
f = open(fdone, 'w')
t = len(points)
f.write("Total Number of Dots annotated are: {}\n".format(t))
if t != 0:
f.write(
"Number of Dots that fall outside of boxes are: {}\n".format(t -
b))
f.write(
"Percentage of the dots falling into the boxes: {}%\n\n".format(
(b / t) * 100))
if segment: