Haven't backed any projects yet!
Generate High-Fidelity Simulated Neutron Detector Data
We have successfully developed a high-fidelity simulation of the Low Intensity Neutron Imaging System (LINIS) using the Geometry and Tracking v4 nuclear physics toolkit via its Python interface, Ge...
Project Update - Materials Cost Adjustment
Thanks again to everyone for helping fully fund this project — work is underway and progressing as planned. Since the proposal was budgeted, the cost of several materials has increased, driven prim...
Background and Methods
In neutron imaging, the goal of source reconstruction is to estimate the angular source intensity distribution S(Ω) from a sparse set of detector counts C_i measured across N_{det} detector element...
PINN Reconstruction Achieves >95% Speed Improvement Over MLEM on Simulated Data
PINN Reconstruction Achieves >95% Speed Improvement Over MLEM on Simulated DataI’m excited to share a key milestone from the project.On simulated datasets, the physics-informed neural network (P...
Expanded neural network's capacity
We've greatly been able to expand our neural network's capacity to localize on source locations by increasing the training data our model uses to localize on test sources. We tested this with anoth...
Proof of concept for the PINN verified
Today we confirmed that our physics-informed neural network can successfully localize a neutron-emitting source while removing the image blurring caused by overlapping detector fields of view. This...