Description:

SpaceNet, obtained via a novel double-stage augmentation framework called FLARE is a hierarchically structured and high-quality astronomical image dataset. It is meticulously designed for both fine-grained and macro classification tasks. Comprising approximately 12,900 samples, SpaceNet incorporates lower (LR) to higher resolution (HR) conversion with standard augmentations and a diffusion approach for synthetic sample generation. This comprehensive dataset enables superior generalization on various recognition tasks, including classification.

Download Dataset

Key Features

High-Resolution Images: The dataset includes high-quality images that facilitate accurate analysis and classification.

Hierarchical Structure: The datajset is hierarchically organized to support both macro and fine-grained classification tasks.

Advanced Augmentation Techniques: Utilizes FLARE framework for double-stage augmentation, enhancing the dataset's diversity and robustness.

Synthetic Sample Generation: Employs a diffusion approach to create synthetic samples, boosting the dataset's size and variability.

Usage

SpaceNet is ideal for:

Training and Evaluation: Developing and testing machine learning models for fine-grained and macro astronomical classification tasks.

Research: Exploring hierarchical classification approaches within the astronomy domain.

Model Development: Creating robust models capable of generalizing across both in-

domain and out-of-domain datasets.

Educational Purposes: Providing a rich dataset for educational projects in astronomy and machine learning.

This dataset is sourced from Kaggle.