AID and Inria. Use kaggle CLI for download.

docs/source/datasets/aid_scene.rst

@@ -0,0 +1,98 @@
+AID Scene Classification
+========================
+
+.. raw:: html
+
+   <p style="display: flex; gap: 10px;">
+   <img src="https://img.shields.io/badge/Task-Scene%20Classification-blue" alt="Task: Scene Classification">
+   <img src="https://img.shields.io/badge/Classes-30-green" alt="Classes: 30">
+   <img src="https://img.shields.io/badge/Domain-Remote%20Sensing-orange" alt="Domain: Remote Sensing">
+   </p>
+
+Overview
+--------
+
+AID (Aerial Image Dataset) is a remote sensing scene classification dataset with 30 scene categories such as airport, port, forest, and residential areas.
+
+This builder downloads the archive from Kaggle and exposes it as a single split:
+
+- **all**: all samples across all 30 classes
+
+Data Structure
+--------------
+
+When accessing an example with ``ds[i]``, the sample contains:
+
+.. list-table::
+   :header-rows: 1
+   :widths: 20 20 60
+
+   * - Key
+     - Type
+     - Description
+   * - ``image``
+     - ``PIL.Image.Image``
+     - RGB aerial scene image
+   * - ``label``
+     - int
+     - Scene class index (0-29)
+
+Usage Example
+-------------
+
+.. code-block:: python
+
+    from stable_datasets.images.aid_scene import AIDScene
+
+    ds = AIDScene(split="all")
+
+    sample = ds[0]
+    print(sample.keys())  # {"image", "label"}
+    print(ds.info.features["label"].names)
+
+Kaggle Requirements
+-------------------
+
+This dataset is downloaded with the **Kaggle CLI**. Install it and configure an API key:
+
+.. code-block:: bash
+
+    pip install kaggle
+
+**Option A — token file (recommended)**
+
+.. code-block:: bash
+
+    mkdir -p ~/.kaggle
+    mv ~/Downloads/kaggle.json ~/.kaggle/kaggle.json
+    chmod 600 ~/.kaggle/kaggle.json
+
+Create ``kaggle.json`` from Kaggle: **Account → API → Create New Token**.
+
+**Option B — environment variables (current shell only)**
+
+.. code-block:: bash
+
+    export KAGGLE_USERNAME="<your_kaggle_username>"
+    export KAGGLE_KEY="<your_api_key>"
+
+References
+----------
+
+- Kaggle dataset: https://www.kaggle.com/datasets/jiayuanchengala/aid-scene-classification-datasets
+
+Citation
+--------
+
+.. code-block:: bibtex
+
+    @article{xia2017aid,
+            title={AID: A benchmark data set for performance evaluation of aerial scene classification},
+            author={Xia, Gui-Song and Hu, Jingwen and Hu, Fan and Shi, Baoguang and Bai, Xiang and Zhong, Yanfei and Zhang, Liangpei and Lu, Xiaoqiang},
+            journal={IEEE Transactions on Geoscience and Remote Sensing},
+            volume={55},
+            number={7},
+            pages={3965--3981},
+            year={2017},
+            publisher={IEEE}
+            }

docs/source/datasets/index.rst

@@ -46,6 +46,7 @@ Available Datasets
 
    arabic_characters
    arabic_digits
+   aid_scene
    cifar10
    cifar100
    cifar10_c
@@ -68,6 +69,7 @@ Available Datasets
    country211
    galaxy10
    hasy_v2
+   inria_aerial_image_labeling
    face_pointing
    rock_paper_scissor
    linnaeus5

docs/source/datasets/inria_aerial_image_labeling.rst

@@ -0,0 +1,112 @@
+Inria Aerial Image Labeling
+===========================
+
+.. raw:: html
+
+   <p style="display: flex; gap: 10px;">
+   <img src="https://img.shields.io/badge/Task-Semantic%20Segmentation-blue" alt="Task: Semantic Segmentation">
+   <img src="https://img.shields.io/badge/Classes-2%20(building%2Fbackground)-green" alt="Classes: 2">
+   <img src="https://img.shields.io/badge/Domain-Remote%20Sensing-orange" alt="Domain: Remote Sensing">
+   </p>
+
+Overview
+--------
+
+The `Inria Aerial Image Labeling`_ benchmark provides high-resolution aerial orthophotos (0.3 m) with pixel-wise labels for **building** vs. **not building**. Training tiles include public ground truth; the official test set is images only (no public masks), so in this builder the **test** split returns ``mask=None``.
+
+Raw archives are downloaded from the Kaggle mirror `sagar100rathod/inria-aerial-image-labeling-dataset` (see `Kaggle dataset`_). Always cite the original paper and follow the dataset license from the official site.
+
+.. _`Inria Aerial Image Labeling`: https://project.inria.fr/aerialimagelabeling/
+.. _`Kaggle dataset`: https://www.kaggle.com/datasets/sagar100rathod/inria-aerial-image-labeling-dataset
+
+Data Structure
+--------------
+
+.. list-table::
+   :header-rows: 1
+   :widths: 20 20 60
+
+   * - Key
+     - Type
+     - Description
+   * - ``image``
+     - ``PIL.Image.Image``
+     - RGB orthophoto (often large GeoTIFF decoded by Pillow)
+   * - ``mask``
+     - ``PIL.Image.Image`` or ``None``
+     - Training: label mask aligned with the image; test: ``None`` (no public GT)
+
+Splits
+------
+
+- **train**: paired ``image`` and ``mask``
+- **test**: ``image`` only if a test image folder is present in the archive; ``mask`` is ``None``
+
+Usage Example
+-------------
+
+.. code-block:: python
+
+    from stable_datasets.images.inria_aerial_image_labeling import InriaAerialImageLabeling
+
+    train_ds = InriaAerialImageLabeling(split="train")
+    sample = train_ds[0]
+    print(sample.keys())  # {"image", "mask"}
+
+    # Optional: load all splits as a dict
+    ds_all = InriaAerialImageLabeling(split=None)
+
+Kaggle setup
+------------
+
+Install the **Kaggle CLI** and configure an **API key** (same pattern as other Kaggle-backed builders):
+
+.. code-block:: bash
+
+    pip install kaggle
+
+**Option A — token file (recommended)**
+
+.. code-block:: bash
+
+    mkdir -p ~/.kaggle
+    mv ~/Downloads/kaggle.json ~/.kaggle/kaggle.json
+    chmod 600 ~/.kaggle/kaggle.json
+
+Create ``kaggle.json`` from Kaggle: **Account → API → Create New Token**.
+
+**Option B — environment variables (current shell only)**
+
+.. code-block:: bash
+
+    export KAGGLE_USERNAME="<your_kaggle_username>"
+    export KAGGLE_KEY="<your_api_key>"
+
+Run the dataset test (downloads a large archive; requires Kaggle auth above):
+
+.. code-block:: bash
+
+    pytest -q -rs stable_datasets/tests/images/test_inria_aerial_image_labeling.py
+
+The test is marked ``large`` because of download size and processing time.
+
+If ``kaggle datasets download`` dies with **SIGKILL** (exit code -9), the process was usually killed by the OS for **low memory** while handling the large zip—not because your API key is wrong. Free RAM, download the archive in a normal terminal first, delete any partial zip, then run the test again.
+
+References
+----------
+
+- Official dataset: https://project.inria.fr/aerialimagelabeling/
+- Kaggle mirror: https://www.kaggle.com/datasets/sagar100rathod/inria-aerial-image-labeling-dataset
+
+Citation
+--------
+
+.. code-block:: bibtex
+
+    @inproceedings{maggiori2017dataset,
+      title={Can Semantic Labeling Methods Generalize to Any City? The Inria Aerial Image Labeling Benchmark},
+      author={Maggiori, Emmanuel and Tarabalka, Yuliya and Charpiat, Guillaume and Alliez, Pierre},
+      booktitle={IEEE International Geoscience and Remote Sensing Symposium (IGARSS)},
+      year={2017},
+      organization={IEEE}
+    }

stable_datasets/images/__init__.py

@@ -2,6 +2,7 @@
 
 from .arabic_characters import ArabicCharacters
 from .arabic_digits import ArabicDigits
+from .aid_scene import AIDScene
 from .awa2 import AWA2
 from .beans import Beans
 from .cars3d import Cars3D
@@ -31,6 +32,7 @@
 from .imagenet_1k import ImageNet1K
 from .imagenet_10 import Imagenette
 from .imagenet_100 import ImageNet100
+from .inria_aerial_image_labeling import InriaAerialImageLabeling
 from .k_mnist import KMNIST
 from .linnaeus5 import Linnaeus5
 from .med_mnist import MedMNIST
@@ -50,6 +52,7 @@
 __all__ = [
     "ArabicCharacters",
     "ArabicDigits",
+    "AIDScene",
     "AWA2",
     "Beans",
     "Cars196",
@@ -77,6 +80,7 @@
     "ImageNet1K",
     "ImageNet100",
     "Imagenette",
+    "InriaAerialImageLabeling",
     "KMNIST",
     "Linnaeus5",
     "MedMNIST",