From 9a49597cef63ea90992a473d53f362db792f9799 Mon Sep 17 00:00:00 2001
From: Jiaqing Zhang <54712081+icey-zhang@users.noreply.github.com>
Date: Tue, 30 Jul 2024 16:33:25 +0800
Subject: [PATCH] Create split_sparcs.py

---
 split_sparcs.py | 131 ++++++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 131 insertions(+)
 create mode 100644 split_sparcs.py

diff --git a/split_sparcs.py b/split_sparcs.py
new file mode 100644
index 0000000..7606cf2
--- /dev/null
+++ b/split_sparcs.py
@@ -0,0 +1,131 @@
+import cv2
+from PIL import Image
+from skimage.io import imread
+from skimage.io import imsave
+import numpy as np
+import os 
+
+#step1: 生成只带有cloud标签的label
+# readpath = "/home/data4/zjq/l8cloudmasks/mask/"
+# list_ri = os.listdir(readpath)
+# savepath = "/home/data4/zjq/l8cloudmasks/label/" #只带有cloud标签的label
+# for ri in list_ri:
+#     label_png = imread(readpath + ri)
+#     label = np.zeros(label_png.shape)
+#     label[np.where(label_png==(255, 255, 255))]=255
+#     lastlabel = label[:,:,0].astype(np.uint8)
+#     imsave(savepath + ri.replace("mask","label"),lastlabel)
+
+#step1: 切分数据集3:1,并且裁剪成200,间隔是200，总共一张1000*1000的图像可以裁剪成25张
+# from PIL import Image
+from skimage import io
+from skimage.util import view_as_windows
+
+# 定义裁剪图像和标签的函数
+def crop_images(image_path, label_path, crop_size=(200, 200), step=200, save_folder=None):
+    image = io.imread(image_path)
+    label = io.imread(label_path)
+    
+    # 确保图像和标签具有相同的尺寸
+    assert image.shape[:2] == label.shape[:2], "Image and label must have the same dimensions"
+    
+    # 计算裁剪的数量
+    num_patches_y = (image.shape[0] - crop_size[1]) // step + 1
+    num_patches_x = (image.shape[1] - crop_size[0]) // step + 1
+    
+    # 裁剪图像和标签
+    for y in range(num_patches_y):
+        for x in range(num_patches_x):
+            # 裁剪图像和标签
+            crop_img = image[y * step:y * step + crop_size[1], x * step:x * step + crop_size[0]]
+            crop_label = label[y * step:y * step + crop_size[1], x * step:x * step + crop_size[0]]
+            
+            # 构造保存的文件名
+            patch_filename = f"{y * num_patches_x + x + 1:01d}_{os.path.basename(image_path)}"
+            io.imsave(os.path.join(save_folder['images'], patch_filename), crop_img)
+            io.imsave(os.path.join(save_folder['labels'], patch_filename.replace('_data.tif', '_label.png')), crop_label)
+
+# 定义保存路径
+base_dir = '/home/data4/zjq/l8cloudmasks'
+train_dir = {
+    'images': os.path.join(base_dir, 'train200v1/images'),
+    'labels': os.path.join(base_dir, 'train200v1/labels')
+}
+test_dir = {
+    'images': os.path.join(base_dir, 'test200v1/images'),
+    'labels': os.path.join(base_dir, 'test200v1/labels')
+}
+
+# 确保保存文件夹存在
+for folder in [train_dir['images'], train_dir['labels'], test_dir['images'], test_dir['labels']]:
+    os.makedirs(folder, exist_ok=True)
+
+# 读取图像和标签文件列表
+img_folder = os.path.join(base_dir, 'img')
+label_folder = os.path.join(base_dir, 'label')
+img_files = [f for f in os.listdir(img_folder) if f.endswith('_data.tif')]
+label_files = [f.replace('_data.tif', '_label.png') for f in img_files]
+
+# 随机划分数据集
+# test_size = 0.25
+# indices = np.arange(len(img_files))
+# np.random.shuffle(indices)
+# split_idx = int(test_size * len(indices))
+# train_indices = indices[split_idx:]
+# test_indices = indices[:split_idx]
+from sklearn.model_selection import train_test_split
+test_size = 0.25
+test_indices = np.zeros(int(test_size * len(img_files)))
+train_indices = np.zeros(int((1-test_size) * len(img_files)))
+# train_indices, test_indices = train_test_split(np.arange(len(img_files)), test_size=test_size, random_state=42)
+test_name = []
+with open("test.txt", 'r', encoding='utf-8') as file:
+    # 逐行读取
+    for line in file:
+        # 处理每一行
+        test_name.append(line.strip())
+        print(line.strip())  # 使用strip()去除可能的前后空白字符
+index_test = 0
+index_train = 0
+for i,label_name in enumerate(label_files):
+    if label_name in test_name:
+        print("test{}:".format(index_test),label_name)
+        test_indices[index_test] = int(i)
+        index_test = index_test + 1
+    else:
+        print("train{}:".format(index_train),label_name)
+        train_indices[index_train] = int(i)
+        index_train = index_train + 1
+
+train_indices = train_indices.astype(int)
+test_indices = test_indices.astype(int)
+# 裁剪图像和标签，并保存到训练集和测试集的文件夹
+for idx in train_indices:
+    img_file = img_files[idx]
+    label_file = label_files[idx]
+    img_path = os.path.join(img_folder, img_file)
+    label_path = os.path.join(label_folder, label_file)
+    
+    # 保存到训练集文件夹
+    save_folder = train_dir
+    crop_images(img_path, label_path, save_folder=save_folder)
+
+for idx in test_indices:
+    img_file = img_files[idx]
+    label_file = label_files[idx]
+    img_path = os.path.join(img_folder, img_file)
+    label_path = os.path.join(label_folder, label_file)
+    
+    # 保存到测试集文件夹
+    save_folder = test_dir
+    crop_images(img_path, label_path, save_folder=save_folder)
+
+
+#类别颜色对应表 
+# 0:0,0,0
+# 1:0,0,128
+# 2:0,0,255
+# 3:0,255,255
+# 4:128,128,128
+# 5:255,255,255 Cloud
+# 6:128,128,0