| from move_file import move_file | |
| from random import shuffle | |
| import numpy as np | |
| import os | |
| from PIL import Image | |
| import shutil | |
| Image.MAX_IMAGE_PIXELS = None | |
| def filter_mask(img_names): | |
| new_img_name = [] | |
| for img_name in img_names: | |
| if 'mask' not in img_name: | |
| new_img_name.append(img_name) | |
| return new_img_name | |
| def checkBlank(patch): | |
| patch = np.array(patch.convert("RGB")) | |
| m = patch.mean() | |
| if 200 <= m <= 255: | |
| return False | |
| else: | |
| return True | |
| def slide_window_inference(image, model): | |
| pass | |
| def slide_crop(img_dataroot,mask_dataroot,img_save_root,mask_save_root): | |
| # clean | |
| if not os.path.exists(img_save_root): | |
| os.mkdir(img_save_root) | |
| os.mkdir(mask_save_root) | |
| else: | |
| shutil.rmtree(img_save_root) | |
| shutil.rmtree(mask_save_root) | |
| os.mkdir(img_save_root) | |
| os.mkdir(mask_save_root) | |
| img_names = os.listdir(img_dataroot) | |
| img_names = filter_mask(img_names) | |
| for img_name in img_names[:]: | |
| # images | |
| img = Image.open(img_dataroot + img_name).convert("RGB") | |
| mask = Image.open(mask_dataroot + img_name[:-4]+'_mask.png') | |
| # params | |
| img_shape = img.size | |
| img_dim = len(img_shape) | |
| window_size = [256, 256] | |
| window_stride = [256, 256] | |
| for d in range(img_dim): | |
| if (window_size[d] is None) or window_size[d] > img_shape[d]: | |
| window_size[d] = img_shape[d] | |
| if (window_stride[d] is None) or window_stride[d] > window_size[d]: | |
| window_stride[d] = window_size[d] | |
| crop_start_list = [] | |
| for w in range(0, img_shape[-1], window_stride[-1]): | |
| w_min = min(w, img_shape[-1] - window_size[-1]) | |
| for h in range(0, img_shape[-2], window_stride[-2]): | |
| h_min = min(h, img_shape[-2] - window_size[-2]) | |
| if img_dim == 2: | |
| crop_start_list.append([h_min, w_min]) | |
| else: | |
| for d in range(0, img_shape[0], window_stride[0]): | |
| d_min = min(d, img_shape[0] - window_size[0]) | |
| crop_start_list.append([d_min, h_min, w_min]) | |
| i = 0 | |
| for c0 in crop_start_list: | |
| c1 = [c0[d] + window_size[d] for d in range(img_dim)] | |
| img_patch = img.crop((c0[0], c0[1], c1[0], c1[1])) | |
| # here, we check the blank patch | |
| if checkBlank(img_patch): | |
| img_patch.save( | |
| img_save_root | |
| + img_name[:-4] | |
| + f"_{i}.jpg" | |
| ) | |
| mask_patch = mask.crop((c0[0], c0[1], c1[0], c1[1])) | |
| mask_patch.save( | |
| mask_save_root | |
| + img_name[:-4] | |
| + f"_{i}_mask.png" | |
| ) | |
| i += 1 | |
| print("done!") | |
| if __name__ == "__main__": | |
| # img_dataroot = "/mnt/data2/lanfz/datasets/digestpath2019/tissue-train-2/images/" | |
| # mask_dataroot = "/mnt/data2/lanfz/datasets/digestpath2019/tissue-train-2/labels_v2/" | |
| # img_save_root = '/mnt/data2/lanfz/datasets/digestpath2019/tissue-train-2-patch/images/' | |
| # mask_save_root = '/mnt/data2/lanfz/datasets/digestpath2019/tissue-train-2-patch/labels_v2/' | |
| # slide_crop(img_dataroot, mask_dataroot, img_save_root, mask_save_root) | |
| img_dataroot = "/mnt/data2/lanfz/datasets/digestpath2019/tissue-train-50/images/" | |
| mask_dataroot = "/mnt/data2/lanfz/datasets/digestpath2019/tissue-train-50/labels/" | |
| img_save_root = '/mnt/data2/lanfz/datasets/digestpath2019/tissue-train-50-patch/images/' | |
| mask_save_root = '/mnt/data2/lanfz/datasets/digestpath2019/tissue-train-50-patch/labels/' | |
| slide_crop(img_dataroot, mask_dataroot, img_save_root, mask_save_root) | |
| # img_dataroot = "/mnt/data2/lanfz/datasets/digestpath2019/tissue-train-10/images/" | |
| # mask_dataroot = "/mnt/data2/lanfz/datasets/digestpath2019/tissue-train-10/labels_v2/" | |
| # img_save_root = '/mnt/data2/lanfz/datasets/digestpath2019/tissue-train-10-patch/images/' | |
| # mask_save_root = '/mnt/data2/lanfz/datasets/digestpath2019/tissue-train-10-patch/labels_v2/' | |
| # slide_crop(img_dataroot, mask_dataroot, img_save_root, mask_save_root) | |
| # img_dataroot = "/mnt/data2/lanfz/datasets/digestpath2019/tissue-train-20/images/" | |
| # mask_dataroot = "/mnt/data2/lanfz/datasets/digestpath2019/tissue-train-20/labels_v2/" | |
| # img_save_root = '/mnt/data2/lanfz/datasets/digestpath2019/tissue-train-20-patch/images/' | |
| # mask_save_root = '/mnt/data2/lanfz/datasets/digestpath2019/tissue-train-20-patch/labels_v2/' | |
| # slide_crop(img_dataroot, mask_dataroot, img_save_root, mask_save_root) | |
| # img_dataroot = "/mnt/data2/lanfz/datasets/digestpath2019/tissue-train-100/images/" | |
| # mask_dataroot = "/mnt/data2/lanfz/datasets/digestpath2019/tissue-train-100/labels_v2/" | |
| # img_save_root = '/mnt/data2/lanfz/datasets/digestpath2019/tissue-train-100-patch/images/' | |
| # mask_save_root = '/mnt/data2/lanfz/datasets/digestpath2019/tissue-train-100-patch/labels_v2/' | |
| # slide_crop(img_dataroot, mask_dataroot, img_save_root, mask_save_root) | |
| # img_dataroot = "/mnt/data2/lanfz/datasets/digestpath2019/tissue-val/images/" | |
| # mask_dataroot = "/mnt/data2/lanfz/datasets/digestpath2019/tissue-val/labels_v2/" | |
| # img_save_root = '/mnt/data2/lanfz/datasets/digestpath2019/tissue-val-patch/images/' | |
| # mask_save_root = '/mnt/data2/lanfz/datasets/digestpath2019/tissue-val-patch/labels_v2/' | |
| # slide_crop(img_dataroot, mask_dataroot, img_save_root, mask_save_root) | |
| # img_dataroot = "/mnt/data2/lanfz/datasets/digestpath2019/tissue-test/images/" | |
| # mask_dataroot = "/mnt/data2/lanfz/datasets/digestpath2019/tissue-test/labels/" | |
| # img_save_root = '/mnt/data2/lanfz/datasets/digestpath2019/tissue-test-patch/images/' | |
| # mask_save_root = '/mnt/data2/lanfz/datasets/digestpath2019/tissue-test-patch/labels/' | |
| # slide_crop(img_dataroot, mask_dataroot, img_save_root, mask_save_root) |