Trans-Neptunian Object Detection
A spatiotemporal U-Net that finds faint moving objects in telescope image sequences.
Mitacs Globalink research at NRC Herzberg, Victoria. Trans-Neptunian Objects are too faint to see in a single exposure, so the network exploits their slow drift across a sequence of difference images. Synthetic objects rendered with real per-frame PSFs are injected into real data for training; the model accepts a variable number of frames and outputs a dense track map. It replaces the classical shift-and-stack search.
Developed with Dr. Sébastien Fabbro at NRC Herzberg.
This repository is private.
The write-up above covers what can be shared publicly. Happy to walk through the code and results on a call, just get in touch.
More in AI / ML
Related projects
AI / ML2023SkinSense: Skin Cancer Classifier
An ensemble of five models labels a lesion as benign or malignant from one RGB image.
- Python
- Scikit-learn
- XGBoost
- LightGBM
AI / ML2025RecycleVision
EfficientNetB0 transfer learning sorts waste into six recyclable categories.
- Python
- TensorFlow
- Keras
- EfficientNetB0
AI / ML2025CIFAR-100 with WideResNet-28×10
A reproducible high-accuracy training pipeline with MixUp, CutMix and RandAugment.
- Python
- PyTorch
- WideResNet
- NumPy