Utilizing machine learning to improve imaging particle detectors
Recent Projects
migYOLO
Published real-time object detection pipeline for the MIGDAL experiment, with automated ML-assisted bounding-box labeling.
OASIS
Segmentation regression framework for fully separating overlapping objects in scientific image data.
MIGDAL_simulator
Fast end-to-end simulation of the MIGDAL detector with a flexible YAML interface.
BEASTnet
2D and 3D convolutional neural networks yielding world-leading directional reconstruction performance in the BEAST TPCs developed at the University of Hawaii.
objectKeyPoint
LabelStudio
General framework to create an object-keypoint detection data labeling interface in LabelStudio, with out-of-the-box support for automated pre-annotations with a pretrained YOLOv8-pose model. Can be adapted to applications outside of physics!