DOTA8 Dataset

Introduction#

Ultralytics DOTA8 is a small oriented object detection dataset composed of the first 8 images of the split DOTAv1 set, 4 for training and 4 for validation. This dataset is ideal for testing and debugging object detection models, or for experimenting with new detection approaches. With 8 images, it is small enough to be easily manageable, yet diverse enough to test training pipelines for errors and act as a sanity check before training larger datasets.

Dataset Structure#

  • Images: 8 aerial tiles (4 train, 4 val) sourced from DOTAv1.

  • Classes: Inherits the 15 DOTAv1 categories such as plane, ship, and large vehicle.

  • Labels: YOLO-format oriented bounding boxes saved as .txt files beside each image.

  • Download: 1 MB, fetched automatically from Ultralytics GitHub assets the first time you train.

  • Recommended layout:

    datasets/dota8/
    ├── images/
    │   ├── train/
    │   └── val/
    └── labels/
        ├── train/
        └── val/

Explore DOTA8 on Ultralytics Platform to browse the aerial tiles with their oriented-box overlays, view the class distribution and annotation heatmaps in the Charts tab, and clone it to train your own model in the cloud.

Dataset YAML#

A YAML file is used to define the dataset configuration. It contains information about the dataset's paths, classes, and other relevant information. In the case of the DOTA8 dataset, the dota8.yaml file is maintained at https://github.com/ultralytics/ultralytics/blob/main/ultralytics/cfg/datasets/dota8.yaml.

ultralytics/cfg/datasets/dota8.yaml
# Ultralytics 🚀 AGPL-3.0 License - https://ultralytics.com/license

# DOTA8 dataset (8 images from the DOTAv1 split) by Ultralytics
# Documentation: https://docs.ultralytics.com/datasets/obb/dota8
# Example usage: yolo train model=yolov8n-obb.pt data=dota8.yaml
# parent
# ├── ultralytics
# └── datasets
#     └── dota8 ← downloads here (1 MB)

# Train/val/test sets as 1) dir: path/to/imgs, 2) file: path/to/imgs.txt, or 3) list: [path/to/imgs1, path/to/imgs2, ..]
path: dota8 # dataset root dir
train: images/train # train images (relative to 'path') 4 images
val: images/val # val images (relative to 'path') 4 images

# Classes for DOTA 1.0
names:
  0: plane
  1: ship
  2: storage tank
  3: baseball diamond
  4: tennis court
  5: basketball court
  6: ground track field
  7: harbor
  8: bridge
  9: large vehicle
  10: small vehicle
  11: helicopter
  12: roundabout
  13: soccer ball field
  14: swimming pool

# Download script/URL (optional)
download: https://github.com/ultralytics/assets/releases/download/v0.0.0/dota8.zip

Usage#

To train a YOLO26n-obb model on the DOTA8 dataset for 100 epochs with an image size of 640, you can use the following code snippets.

Train Example
from ultralytics import YOLO

# Load a model
model = YOLO("yolo26n-obb.pt")  # load a pretrained model (recommended for training)

# Train the model
results = model.train(data="dota8.yaml", epochs=100, imgsz=640)

Sample Images and Annotations#

Here are some examples of images from the DOTA8 dataset, along with their corresponding annotations:

DOTA8 oriented bounding box dataset training mosaic
  • Mosaiced Image: This image demonstrates a training batch composed of mosaiced dataset images. Mosaicing is a technique used during training that combines multiple images into a single image to increase the variety of objects and scenes within each training batch. This helps improve the model's ability to generalize to different object sizes, aspect ratios, and contexts.

The example showcases the variety and complexity of the images in the DOTA8 dataset and the benefits of using mosaicing during the training process.

Citations and Acknowledgments#

If you use the DOTA dataset in your research or development work, please cite the following paper:

Citation
@article{9560031,
  author={Ding, Jian and Xue, Nan and Xia, Gui-Song and Bai, Xiang and Yang, Wen and Yang, Michael and Belongie, Serge and Luo, Jiebo and Datcu, Mihai and Pelillo, Marcello and Zhang, Liangpei},
  journal={IEEE Transactions on Pattern Analysis and Machine Intelligence},
  title={Object Detection in Aerial Images: A Large-Scale Benchmark and Challenges},
  year={2022},
  volume={44},
  number={11},
  pages={7778-7796},
  doi={10.1109/TPAMI.2021.3117983}
}

We thank the team behind the DOTA datasets for curating them. For full details on the dataset, visit the official DOTA website.

FAQ#

  • The DOTA8 dataset is a small oriented object detection dataset made up of the first 8 images from the DOTAv1 split set, with 4 images designated for training and 4 for validation. It's ideal for testing and debugging object detection models like Ultralytics YOLO26. Due to its manageable size and diversity, it helps in identifying pipeline errors and running sanity checks before training on larger datasets.

  • To train a YOLO26n-obb model on the DOTA8 dataset for 100 epochs with an image size of 640, you can use the following code snippets. For comprehensive argument options, refer to the model Training page.

    Train Example
    from ultralytics import YOLO
    
    # Load a model
    model = YOLO("yolo26n-obb.pt")  # load a pretrained model (recommended for training)
    
    # Train the model
    results = model.train(data="dota8.yaml", epochs=100, imgsz=640)
  • The DOTA dataset is known for its large-scale benchmark and the challenges it presents for object detection in aerial images. The DOTA8 subset is a smaller, manageable dataset ideal for initial tests. The dota8.yaml configuration file, which contains paths, classes, and configuration details, is maintained in the Ultralytics GitHub repository.

  • Mosaicing combines multiple images into one during training, increasing the variety of objects and contexts within each batch. This improves a model's ability to generalize to different object sizes, aspect ratios, and scenes.

  • Ultralytics YOLO26 supports real-time object detection, including oriented bounding boxes (OBB), alongside other tasks such as instance segmentation, all through one training pipeline. Its pretrained models make fine-tuning efficient.

Comments2

Ask a question, share your results, or help someone else.

  1. expen6veGitHub

    How do you create a dataset for OBB? can i use the visdrone dataset to train an obb model?

    1. Glenn JocherUltralytics

      @expen6ve yes, you can use VisDrone with YOLO OBB by converting your annotations to YOLO OBB format (four corner points with normalized coordinates). Create a dataset YAML file following our DOTA8 example structure, and ensure annotations match the required class_index x1 y1 x2 y2 x3 y3 x4 y4 format. For conversion tools and format details, see our OBB Dataset Guide.