JWST Astronomical Object Detection Model

This is a fine-tuned YOLO model specifically trained for detecting astronomical objects in JWST (James Webb Space Telescope) images.

Model Details

Architecture: YOLOv8n (nano)
Training Data: 2,587 high-quality JWST images
Classes: 2 (bright_object, galaxy_like)
Performance: 26.7% mAP50, 52.7% precision on bright objects
Training Time: 75 epochs (~25 hours)

Usage

from ultralytics import YOLO

# Load the model
model = YOLO("norbertm/jwst-astronomical-detection")

# Run inference
results = model("path/to/jwst/image.png", conf=0.15)

Training Details

Dataset: 2,587 JWST images with automated annotations
Instruments: NIRCAM (Near-Infrared Camera)
Filters: F090W, F150W, F200W, F277W, F356W, F444W
Targets: Stephan's Quintet, M16, NGC 3324, NGC 3132, SMACS J0723.3-7327, WASP-39b

Research Applications

Automated astronomical object detection
Multi-wavelength object correlation
Quality assessment of JWST data
Large-scale astronomical surveys

Citation

If you use this model in your research, please cite:

@dataset{jwst_quality_analysis,
  title={JWST Quality Analysis Dataset},
  author={Your Name},
  year={2024},
  url={https://huggingface.co/datasets/norbertm/jwst-quality-analysis-dataset}
}

License

MIT License - see LICENSE file for details.

Downloads last month: 2

norbertm
/

jwst-astronomical-detection