Update OSCD training data

tests/data/README.md

@@ -53,18 +53,28 @@ VisionDataset data can be created like so.
 ### RGB images
 
 ```python
+import numpy as np
 from PIL import Image
 
-img = Image.new("RGB", (64, 64))
+DTYPE = np.uint8
+SIZE = 64
+
+arr = np.random.randint(np.iinfo(DTYPE).max, size=(SIZE, SIZE, 3), dtype=DTYPE)
+img = Image.fromarray(arr)
 img.save("01.png")
 ```
 
 ### Grayscale images
 
 ```python
+import numpy as np
 from PIL import Image
 
-img = Image.new("L", (64, 64))
+DTYPE = np.uint8
+SIZE = 64
+
+arr = np.random.randint(np.iinfo(DTYPE).max, size=(SIZE, SIZE), dtype=DTYPE)
+img = Image.fromarray(arr)
 img.save("02.jpg")
 ```
 
@@ -84,9 +94,12 @@ wavfile.write("01.wav", rate=22050, data=audio)
 import h5py
 import numpy as np
 
-num_classes = 10
-images = np.random.randint(256, size=(64, 64, 3), dtype=np.uint8)
-masks = np.random.randint(num_classes, size=(64, 64), dtype=np.uint8)
+DTYPE = np.uint8
+SIZE = 64
+NUM_CLASSES = 10
+
+images = np.random.randint(np.iinfo(DTYPE).max, size=(SIZE, SIZE, 3), dtype=DTYPE)
+masks = np.random.randint(NUM_CLASSES, size=(SIZE, SIZE), dtype=DTYPE)
 with h5py.File("data.hdf5", "w") as f:
     f.create_dataset("images", data=images)
     f.create_dataset("masks", data=masks)

tests/data/oscd/Onera Satellite Change Detection dataset - Images/test/dates.txt

@@ -1,2 +1,2 @@
-date_1: 20151211
-date_2: 20180330
+date_1: 20161130
+date_2: 20170829

tests/data/oscd/data.py

@@ -0,0 +1,84 @@
+#!/usr/bin/env python3
+
+# Copyright (c) Microsoft Corporation. All rights reserved.
+# Licensed under the MIT License.
+
+import hashlib
+import os
+import shutil
+
+import numpy as np
+from PIL import Image
+
+
+SIZE = 64  # image width/height
+
+np.random.seed(0)
+
+directories = [
+    "Onera Satellite Change Detection dataset - Images",
+    "Onera Satellite Change Detection dataset - Train Labels",
+    "Onera Satellite Change Detection dataset - Test Labels",
+]
+bands = [
+    "B01",
+    "B02",
+    "B03",
+    "B04",
+    "B05",
+    "B06",
+    "B07",
+    "B08",
+    "B09",
+    "B10",
+    "B11",
+    "B12",
+    "B8A",
+]
+
+# Remove old data
+for directory in directories:
+    filename = f"{directory}.zip"
+
+    if os.path.exists(filename):
+        os.remove(filename)
+    if os.path.exists(directory):
+        shutil.rmtree(directory)
+
+# Create images
+for subdir in ["train1", "train2", "test"]:
+    for rect in ["imgs_1_rect", "imgs_2_rect"]:
+        directory = os.path.join(directories[0], subdir, rect)
+        os.makedirs(directory)
+
+        for band in bands:
+            filename = os.path.join(directory, f"{band}.tif")
+            arr = np.random.randint(
+                np.iinfo(np.uint16).max, size=(SIZE, SIZE), dtype=np.uint16
+            )
+            img = Image.fromarray(arr)
+            img.save(filename)
+
+    filename = os.path.join(directories[0], subdir, "dates.txt")
+    with open(filename, "w") as f:
+        for key, value in [("date_1", "20161130"), ("date_2", "20170829")]:
+            f.write(f"{key}: {value}\n")
+
+# Create labels
+for i, subdir in [(1, "train1"), (1, "train2"), (2, "test")]:
+    directory = os.path.join(directories[i], subdir, "cm")
+    os.makedirs(directory)
+    filename = os.path.join(directory, "cm.png")
+    arr = np.random.randint(np.iinfo(np.uint8).max, size=(SIZE, SIZE), dtype=np.uint8)
+    img = Image.fromarray(arr)
+    img.save(filename)
+
+for directory in directories:
+    # Compress data
+    shutil.make_archive(directory, "zip", ".", directory)
+
+    # Compute checksums
+    filename = f"{directory}.zip"
+    with open(filename, "rb") as f:
+        md5 = hashlib.md5(f.read()).hexdigest()
+        print(repr(filename) + ": " + repr(md5) + ",")