final update

README.md

@@ -22,6 +22,7 @@ pip install git+https://github.com/isaaccorley/torchrs
 * [PROBA-V Super Resolution](https://github.com/isaaccorley/torchrs#proba-v-super-resolution)
 * [ETCI 2021 Flood Detection](https://github.com/isaaccorley/torchrs#etci-2021-flood-detection)
 * [Onera Satellite Change Detection (OSCD)](https://github.com/isaaccorley/torchrs#onera-satellite-change-detection-oscd)
+* [Satellite Side-Looking (S2Looking) Change Detection](https://github.com/isaaccorley/torchrs#satellite-side-looking-s2looking-change-detection)
 * [LEVIR Change Detection+ (LEVIR-CD+)](https://github.com/isaaccorley/torchrs#levir-change-detection-levir-cd-)
 * [Remote Sensing Visual Question Answering (RSVQA) Low Resolution (LR)](https://github.com/isaaccorley/torchrs#remote-sensing-visual-question-answering-rsvqa-low-resolution-lr)
 * [Remote Sensing Image Captioning Dataset (RSICD)](https://github.com/isaaccorley/torchrs#remote-sensing-image-captioning-dataset-rsicd)
@@ -122,11 +123,42 @@ x: dict(
 """
 ```
 
+
+### Satellite Side-Looking (S2Looking) Change Detection
+
+<img src="./assets/s2looking.png" width="500px"></img>
+
+The [S2Looking](https://github.com/AnonymousForACMMM/Dataset) dataset, proposed in ["S2Looking: A Satellite Side-Looking Dataset for Building Change Detection", Shen et al.](https://arxiv.org/abs/2107.09244) is a rural building Change Detection dataset of 5,000 very high resolution (VHR) 0.5-0.8m registered RGB image pairs of varying off-nadir angles taken by the (Gaogen (GF))[https://earth.esa.int/web/eoportal/satellite-missions/g/gaofen-1], (SuperView (SV))[https://eos.com/find-satellite/superview-1/], and (BeiJing-2 (BJ-2))[https://space.oscar.wmo.int/satelliteprogrammes/view/beijing_2] satellites. The dataset contains separate new and demolished building masks from regions all over the Earth a time span of 1-3 years. This dataset was proposed along with the LEVIR-CD+ dataset and is considered difficult due to the rural locations and off-nadir angles.
+
+The dataset can be downloaded (11GB) using `scripts/download_s2looking.sh` and instantiated below:
+
+```python
+from torchrs.transforms import Compose, ToTensor
+from torchrs.datasets import S2Looking
+
+transform = Compose([ToTensor()])
+
+dataset = S2Looking(
+    root="path/to/dataset/",
+    split="train",  # or 'val', 'test'
+    transform=transform,
+)
+
+x = dataset[0]
+"""
+x: dict(
+    x: (2, 3, 1024, 1024)
+    build_mask: (1, 1024, 1024),
+    demolish_mask: (1, 1024, 1024)
+)
+"""
+```
+
 ### LEVIR Change Detection+ (LEVIR-CD+)
 
 <img src="./assets/levircd_plus.png" width="600px"></img>
 
-The [LEVIR-CD+](https://github.com/AnonymousForACMMM/Dataset) dataset, proposed in ["S2Looking: A Satellite Side-Looking Dataset for Building Change Detection", Shen et al.](https://arxiv.org/abs/2107.09244) is an urban Change Detection dataset of 985 very high resolution (VHR) 0.5m RGB image pairs extracted from Google Earth. The dataset contains building/land use change masks from 20 different regions of Texas between 2002-2020 with a temporal difference of 5 years.
+The [LEVIR-CD+](https://github.com/AnonymousForACMMM/Dataset) dataset, proposed in ["S2Looking: A Satellite Side-Looking Dataset for Building Change Detection", Shen et al.](https://arxiv.org/abs/2107.09244) is an urban building Change Detection dataset of 985 very high resolution (VHR) 0.5m RGB image pairs extracted from Google Earth. The dataset contains building/land use change masks from 20 different regions of Texas between 2002-2020 with a time span of 5 years. This dataset was proposed along with the S2Looking dataset and is considered the easier version due to the urban locations and near-nadir angles.
 
 The dataset can be downloaded (3.6GB) using `scripts/download_levircd_plus.sh` and instantiated below:
 

torchrs/datasets/__init__.py

@@ -5,10 +5,11 @@
 from .resisc45 import RESISC45
 from .rsicd import RSICD
 from .oscd import OSCD
+from .s2looking import S2Looking
 from .levircd import LEVIRCD_Plus
 
 
 __all__ = [
     "PROBAV", "ETCI2021", "RSVQALR", "EuroSATRGB", "EuroSATMS",
-    "RESISC45", "RSICD", "OSCD", "LEVIRCD_Plus"
+    "RESISC45", "RSICD", "OSCD", "S2Looking", "LEVIRCD_Plus"
 ]

torchrs/datasets/levircd.py

@@ -48,7 +48,7 @@ def __len__(self) -> int:
         return len(self.files)
 
     def __getitem__(self, idx: int) -> Dict:
-        """ Returns a dict containing x, mask, region, dates
+        """ Returns a dict containing x, mask
         x: (2, 13, h, w)
         mask: (1, h, w)
         """

torchrs/datasets/oscd.py

@@ -67,7 +67,7 @@ def __len__(self) -> int:
         return len(self.regions)
 
     def __getitem__(self, idx: int) -> Dict:
-        """ Returns a dict containing x, mask, region, dates
+        """ Returns a dict containing x, mask
         x: (2, 13, h, w)
         mask: (1, h, w)
         """

torchrs/datasets/s2looking.py

@@ -3,7 +3,6 @@
 from typing import Dict
 
 import torch
-import tifffile
 import numpy as np
 from PIL import Image
 
@@ -29,44 +28,40 @@ def __init__(
         split: str = "train",
         transform: Compose = Compose([ToTensor()]),
     ):
-        assert split in ["train", "test"]
+        assert split in ["train", "val", "test"]
         self.root = root
         self.transform = transform
         self.files = self.load_files(root, split)
 
     @staticmethod
     def load_files(root: str, split: str):
-        images = []
-        labels_root = os.path.join(root, f"{split}_labels")
-        images_root = os.path.join(root, "images")
-        folders = glob(os.path.join(labels_root, "*/"))
-        for folder in folders:
-            region = folder.split(os.sep)[-2]
-            mask = os.path.join(labels_root, region, "cm", "cm.png")
-            images1 = glob(os.path.join(images_root, region, "imgs_1_rect", "*.tif"))
-            images2 = glob(os.path.join(images_root, region, "imgs_2_rect", "*.tif"))
-            images1 = sorted(images1, key=sort)
-            images2 = sorted(images2, key=sort)
-            with open(os.path.join(images_root, region, "dates.txt")) as f:
-                dates = tuple([line.split()[-1] for line in f.read().strip().splitlines()])
-
-            regions.append(dict(region=region, images1=images1, images2=images2, mask=mask, dates=dates))
-
-        return regions
+        files = []
+        images = glob(os.path.join(root, split, "Image1", "*.png"))
+        images = sorted([os.path.basename(image) for image in images])
+        for image in images:
+            image1 = os.path.join(root, split, "Image1", image)
+            image2 = os.path.join(root, split, "Image2", image)
+            build_mask = os.path.join(root, split, "label1", image)
+            demo_mask = os.path.join(root, split, "label2", image)
+            files.append(dict(image1=image1, image2=image2, build_mask=build_mask, demolish_mask=demo_mask))
+        return files
 
     def __len__(self) -> int:
-        return len(self.regions)
+        return len(self.files)
 
     def __getitem__(self, idx: int) -> Dict:
-        """ Returns a dict containing x, mask, region, dates
+        """ Returns a dict containing x, mask
         x: (2, 13, h, w)
-        mask: (1, h, w)
+        build_mask: (1, h, w)
+        demolish_mask: (1, h, w)
         """
-        region = self.regions[idx]
-        mask = np.array(Image.open(region["mask"]))
-        mask[mask == 255] = 1
-        image1 = np.stack([tifffile.imread(path) for path in region["images1"]], axis=0)
-        image2 = np.stack([tifffile.imread(path) for path in region["images2"]], axis=0)
-        image1, image2, mask = self.transform([image1, image2, mask])
+        files = self.files[idx]
+        build_mask = np.array(Image.open(files["build_mask"]))
+        demo_mask = np.array(Image.open(files["demolish_mask"]))
+        build_mask = np.clip(build_mask.mean(axis=-1), 0, 1).astype("uint8")
+        demo_mask = np.clip(demo_mask.mean(axis=-1), 0, 1).astype("uint8")
+        image1 = np.array(Image.open(files["image1"]))
+        image2 = np.array(Image.open(files["image2"]))
+        image1, image2, build_mask, demo_mask = self.transform([image1, image2, build_mask, demo_mask])
         x = torch.stack([image1, image2], dim=0)
-        return dict(x=x, mask=mask)
+        return dict(x=x, build_mask=build_mask, demolish_mask=demo_mask)