Stardist简单使用

官方文档里含几个例子
用于细胞分割，基于unet

pip install stardist

from csbdeep.utils import Path, download_and_extract_zip_file, normalize
from stardist.matching import matching_dataset
from stardist import fill_label_holes, random_label_cmap, relabel_image_stardist, calculate_extents, gputools_available, _draw_polygons
from stardist.models import Config2D, StarDist2D, StarDistData2D

#读图
fX = sorted(Path('路径/').glob('*.tif'))
fY = sorted(Path('路径').glob('*.tif'))
# assert all(Path(x).name==Path(y).name for x,y in zip(fX,fY))
print(f"{len(fX)} files found")

#预处理，会要求灰度图
X = list(map(imread,map(str,tqdm(fX))))
Y = list(map(imread,map(str,tqdm(fY))))
n_channel = 1 if X[0].ndim == 2 else X[0].shape[-1]

axis_norm = (0,1)   # normalize channels independently
# axis_norm = (0,1,2) # normalize channels jointly
if n_channel > 1:
    print("Normalizing image channels %s." % ('jointly' if axis_norm is None or 2 in axis_norm else 'independently'))
    sys.stdout.flush()

X = [normalize(x,1,99.8,axis=axis_norm) for x in tqdm(X)]
Y = [fill_label_holes(y) for y in tqdm(Y)]

#区分训练集
# assert len(X) > 1, "not enough training data"
rng = np.random.RandomState(42)
ind = rng.permutation(len(X))
n_val = max(1, int(round(0.15 * len(ind))))
ind_train, ind_val = ind[:-n_val], ind[-n_val:]
X_val, Y_val = [X[i] for i in ind_val]  , [Y[i] for i in ind_val]
X_trn, Y_trn = [X[i] for i in ind_train], [Y[i] for i in ind_train] 
print('number of images: %3d' % len(X))
print('- training:       %3d' % len(X_trn))
print('- validation:     %3d' % len(X_val))

#数据增广
def random_fliprot(img, mask): 
    assert img.ndim >= mask.ndim
    axes = tuple(range(mask.ndim))
    perm = tuple(np.random.permutation(axes))
    img = img.transpose(perm + tuple(range(mask.ndim, img.ndim))) 
    mask = mask.transpose(perm) 
    for ax in axes: 
        if np.random.rand() > 0.5:
            img = np.flip(img, axis=ax)
            mask = np.flip(mask, axis=ax)
    return img, mask 

def random_intensity_change(img):
    img = img*np.random.uniform(0.6,2) + np.random.uniform(-.2,.2)
    return img


def augmenter(img,mask):
    """Augmentation for image,mask"""
    img, mask = random_fliprot(img, mask)
    img = random_intensity_change(img)
    return img, mask

#预训练模型
StarDist2D.from_pretrained()
model = StarDist2D.from_pretrained('2D_versatile_fluo')

#自定义模型
my_model= StarDist2D(config ,name='my_model', basedir='存储路径')

#训练
my_model.train(X_trn, Y_trn, validation_data=(X_val,Y_val), augmenter=augmenter,epochs=200, steps_per_epoch=25)

#预测
mylabel, _ = my_model.predict_instances(normalize(a))

#再使用
my_model = StarDist2D(None, name='my_model', basedir='存储路径')