Amplifying the imaging power of ground-based telescopes with space-based telescopes data and generative AI.
This project facilitates the download of ground-based telescope imagery based on celestial coordinates (RA/Dec), converts it into a usable image format, and enhances it by utilizing contextual information from space-based telescopes using a custom-trained Conditional Generative Adversarial Network (cGAN).
The cGAN model was trained on paired images from ground-based and space-based telescopes, with the goal of transforming ground-based observations to match the quality and clarity of space-based imagery.
- 🔭 Input celestial coordinates (RA, Dec)
- 📡 Download raw FITS files from public astronomical databases
- 🖼️ Convert FITS to 8-bit TIF format
- 🤖 Enhance the image using a pre-trained cGAN (Conditional GAN) model
- 💾 Save the enhanced image in
.pngformat
Here are some sample outputs from our model, demonstrating its ability to enhance ground-based images to a quality that closely resembles space-based observations. Impressively, the model maintains high performance even when the input lacks clear visual details of the galaxy.
The cGAN model, train dataset and a catalog of 63,202 enhanced images are available at: https://doi.org/10.6084/m9.figshare.30226591
git clone https://github.com/SaiTeja-Erukude/Enhancing-Ground-Based-Astronomy-using-GenAI.git
cd galaxy-enhancerpython3.8 -m venv .venv
.venv\Scripts\activate # For Windowspip install -r requirements.txtpython main.py