added patternnet dataset, datamodule, and download script. updated re…

…adme

README.md

@@ -57,6 +57,7 @@ pip install 'git+https://github.com/isaaccorley/torchrs.git#egg=torch-rs[train]'
 * [GID-15 - Semantic Segmentation](https://github.com/isaaccorley/torchrs#gid-15)
 * [TiSeLaC - Time-Series Land Cover Classification](https://github.com/isaaccorley/torchrs#tiselac)
 * [UC Merced - Land Use Classification](https://github.com/isaaccorley/torchrs#uc-merced-ucm)
+* [PatternNet - Image Retrieval / Scene Classification](https://github.com/isaaccorley/torchrs#patternnet)
 
 ### PROBA-V Super Resolution
 
@@ -886,6 +887,42 @@ dataset.classes
 """
 ```
 
+### PatternNet
+
+<img src="./assets/patternnet.png" width="500px"></img>
+
+The [PatternNet](https://sites.google.com/view/zhouwx/dataset) dataset, proposed in ["PatternNet: A Benchmark Dataset for Performance Evaluation of Remote Sensing Image Retrieval", Yang et al.](https://arxiv.org/abs/1706.03424) is a image retrieval and scene classification dataset of 30,400 256x256 high resolution (.06-5m) RGB images extracted using [Google Earth](https://earth.google.com/web/) and [Google Maps](https://developers.google.com/maps) with 38 scene classes (800 images per class). This dataset was originally proposed as a remote sensing image retrieval (RSIR) dataset with classes selected for high intra-class diversity and inter-class similarity such that image retrieval requires learning fine-grained details between multiple classes. Additionally, this dataset has some unique classes not found in other scene classification datasets, e.g. oil well, nursing home, solar panel, etc.
+
+The dataset can be downloaded (1.4GB) using `scripts/download_patternnet.sh` and instantiated below:
+
+```python
+import torchvision.transforms as T
+from torchrs.datasets import PatternNet
+
+transform = T.Compose([T.ToTensor()])
+
+dataset = PatternNet(
+    root="path/to/dataset/",
+    transform=transform
+)
+
+x, y = dataset[0]
+"""
+x: (3, 256, 256)
+y: int
+"""
+
+dataset.classes
+"""
+['airplane', 'baseball_field', 'basketball_court', 'beach', 'bridge', 'cemetery', 'chaparral',
+'christmas_tree_farm', 'closed_road', 'coastal_mansion', 'crosswalk', 'dense_residential',
+'ferry_terminal', 'football_field', 'forest', 'freeway', 'golf_course', 'harbor', 'intersection',
+'mobile_home_park', 'nursing_home', 'oil_gas_field', 'oil_well', 'overpass', 'parking_lot', 'parking_space',
+'railway', 'river', 'runway', 'runway_marking', 'shipping_yard', 'solar_panel','sparse_residential',
+'storage_tank', 'swimming_pool', 'tennis_court', 'transformer_station', 'wastewater_treatment_plant']
+"""
+```
+
 ## Models
 
 * [Multi-Image Super Resolution - RAMS](https://github.com/isaaccorley/torchrs#multi-image-super-resolution---rams)

scripts/download_patternnet.sh

@@ -0,0 +1,5 @@
+pip install gdown
+mkdir -p .data/
+gdown --id 127lxXYqzO6Bd0yZhvEbgIfz95HaEnr9K
+unzip PatternNet.zip -d .data/
+rm PatternNet.zip

torchrs/datasets/__init__.py

@@ -23,12 +23,13 @@
 from .dubai_segmentation import DubaiSegmentation
 from .hkh_glacier import HKHGlacierMapping
 from .ucm import UCM
+from .patternnet import PatternNet
 
 
 __all__ = [
     "PROBAV", "ETCI2021", "RSVQALR", "RSVQAxBEN", "EuroSATRGB", "EuroSATMS",
     "RESISC45", "RSICD", "OSCD", "S2Looking", "LEVIRCDPlus", "FAIR1M",
     "SydneyCaptions", "UCMCaptions", "S2MTCP", "ADVANCE", "SAT4", "SAT6",
     "HRSCD", "InriaAIL", "Tiselac", "GID15", "ZueriCrop", "AID", "DubaiSegmentation",
-    "HKHGlacierMapping", "UCM"
+    "HKHGlacierMapping", "UCM", "PatternNet"
 ]

torchrs/datasets/patternnet.py

@@ -0,0 +1,21 @@
+import os
+
+import torchvision.transforms as T
+from torchvision.datasets import ImageFolder
+
+
+class PatternNet(ImageFolder):
+    """ PatternNet dataset from 'PatternNet: A benchmark dataset for performance
+    evaluation of remote sensing image retrieval', Zhou at al. (2018)
+    https://arxiv.org/abs/1706.03424
+
+    """
+    def __init__(
+        self,
+        root: str = ".data/PatternNet",
+        transform: T.Compose = T.Compose([T.ToTensor()])
+    ):
+        super().__init__(
+            root=os.path.join(root, "images"),
+            transform=transform
+        )

torchrs/train/datamodules/__init__.py

@@ -23,6 +23,7 @@
 from .dubai_segmentation import DubaiSegmentationDataModule
 from .hkh_glacier import HKHGlacierMappingDataModule
 from .ucm import UCMDataModule
+from .patternnet import PatternNetDataModule
 
 
 __all__ = [
@@ -32,5 +33,5 @@
     "FAIR1MDataModule", "SydneyCaptionsDataModule", "UCMCaptionsDataModule", "S2MTCPDataModule",
     "ADVANCEDataModule", "SAT4DataModule", "SAT6DataModule", "HRSCDDataModule", "InriaAILDataModule",
     "TiselacDataModule", "GID15DataModule", "ZueriCropDataModule", "AIDDataModule",
-    "DubaiSegmentationDataModule", "HKHGlacierMappingDataModule", "UCMDataModule"
+    "DubaiSegmentationDataModule", "HKHGlacierMappingDataModule", "UCMDataModule", "PatternNetDataModule"
 ]

torchrs/train/datamodules/patternnet.py

@@ -0,0 +1,26 @@
+from typing import Optional
+
+import torchvision.transforms as T
+
+from torchrs.datasets.utils import dataset_split
+from torchrs.train.datamodules import BaseDataModule
+from torchrs.datasets import PatternNet
+
+
+class PatternNetDataModule(BaseDataModule):
+
+    def __init__(
+        self,
+        root: str = ".data/PatternNet",
+        transform: T.Compose = T.Compose([T.ToTensor()]),
+        *args, **kwargs
+    ):
+        super().__init__(*args, **kwargs)
+        self.root = root
+        self.transform = transform
+
+    def setup(self, stage: Optional[str] = None):
+        dataset = PatternNet(root=self.root, transform=self.transform)
+        self.train_dataset, self.val_dataset, self.test_dataset = dataset_split(
+            dataset, val_pct=self.val_split, test_pct=self.test_split
+        )