metadata

dataset_info:
  features:
    - name: image_id
      dtype: int64
    - name: image
      dtype: image
    - name: width
      dtype: int64
    - name: height
      dtype: int64
    - name: objects
      sequence:
        - name: bbox_id
          dtype: int64
        - name: category
          dtype:
            class_label:
              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': small-vehicle
                '10': large-vehicle
                '11': roundabout
                '12': swimming-pool
                '13': helicopter
                '14': soccer-ball-field
                '15': container-crane
        - name: bbox
          sequence: int64
          length: 4
        - name: area
          dtype: int64
  splits:
    - name: train
      num_bytes: 5043231102.186
      num_examples: 54087
    - name: validation
      num_bytes: 184865300
      num_examples: 2000
    - name: test
      num_bytes: 628863995.564
      num_examples: 6854
  download_size: 5791640499
  dataset_size: 5856960397.75
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
      - split: validation
        path: data/validation-*
      - split: test
        path: data/test-*
license: mit
task_categories:
  - object-detection
language:
  - en
pretty_name: DOTA COCO format

DOTA: Resized and Hugging Face-Ready Vision Dataset

This dataset is a restructured version of the DOTA (Dataset for Object Detection in Aerial Images), specifically designed to simplify object detection workflows. By resizing the original images and converting them to the COCO format, this project provides an easier way to use DOTA with popular computer vision frameworks. Additionally, the dataset is formatted for seamless integration with Hugging Face datasets, unlocking new possibilities for training and experimentation.

📂 Dataset Structure

COCO Format

The dataset follows the COCO dataset structure, making it straightforward to work with:

dota/  
├── annotations/  
│   ├── instances_train.json  
│   ├── instances_val.json  
│   └── instances_test.json  
├── train/
├── val/
├── test/

Hugging Face Format

The dataset is compatible with the datasets library. You can load it directly using:

from datasets import load_dataset  

dataset = load_dataset("HichTala/dota")

🖼️ Sample Visualizations

Above: An example of resized images with bounding boxes in COCO format.

🚀 Getting Started

Install Required Libraries

Install datasets for Hugging Face compatibility:

pip install datasets

Use any object detection framework supporting COCO format for training.

Load the Dataset

Hugging Face:

from datasets import load_dataset  

dataset = load_dataset("HichTala/dota")  
train_data = dataset["train"]

Custom Script for COCO-Compatible Frameworks:

import json  
from pycocotools.coco import COCO

coco = COCO("annotations/train.json")

see demo notebook here for more details.

⚙️ Preprocessing Details

Resizing: The original large images were split into smaller patches, each resized to 512x512 pixels.
Annotations: Converted to COCO format for better compatibility and flexibility.

📝 How to Cite

If you use this dataset, please consider citing the original DOTA dataset:

Copy code
@inproceedings{Xia_2018_CVPR,  
    author = {Gui-Song Xia and Xiang Bai and Jieqing Zheng and others},  
    title = {DOTA: A Large-Scale Dataset for Object Detection in Aerial Images},  
    booktitle = {CVPR},  
    year = {2018}  
}

Additionally, you can mention this repository for the resized COCO and Hugging Face formats.

Enjoy using DOTA in coco format for your object detection experiments! 🚀