Companion code of journal articles
L. Lozenski, M. Anastasio, U. Villa. A Memory-Efficient Self-Supervised Dynamic Image Reconstruction Method using Neural Fields, IEEE Transactions on Computational Imaging 8 (2022): 879-892. (preprint)
L. Lozenski, R. Cam, M. Pagel, M. Anastasio. ProxNF: Neural Field Proximal Training for High-Resolution 4D Dynamic Image Reconstruction, submitted to IEEE Transactions on Computational Imaging. (preprint)
Neural Fields for solving dynamic CRT imaging problems
Neural fields are a particular class of neural networks representing the dynamic object as a continuous function of space and time. Neural field representation reduces image reconstruction to estimate the network parameters via a nonlinear optimization problem (training). Once trained, the neural field can be evaluated at arbitrary locations in space and time, allowing for high-resolution rendering of the object. Key advantages of the proposed approach are that neural fields automatically learn and exploit redundancies in the sought-after object to both regularize the reconstruction and significantly reduce memory storage requirements.
In this repository, we display this proposed neural field framework with a supervised learning example and two unsupervised image reconstruction examples using the dynamic circular radon transform (CRT).
PyTorch: open source machine learning framework that accelerates the path from research prototyping to production deployment.
conda install pytorch torchvision torchaudio -c pytorchscikit-image: collection of algorithms for image processing
conda install scikit-image