BioMedAI-UCSC/InverseSR
BioMedAI-UCSC/InverseSR has 77 GitHub stars and 6 forks. GitHot measured 0 stars in 24 hours and 0 in 7 days. Last checked 2026-11-20 09:48 UTC. Last push 2026-08-25.
Crecimiento de estrellas medido
- in 24h
- 0
- in 7 days
- 0
- in 15 days
- Recopilando historial
- in 1 month
- Recopilando historial
Cambios netos en períodos que se solapan, no totales diarios. Los datos no disponibles se dejan en blanco. Las estrellas indican interés, no calidad.
Acerca de
[Early Accepted at MICCAI 2023] Pytorch Code of "InverseSR: 3D Brain MRI Super-Resolution Using a Latent Diffusion Model"
Aprendizaje automáticoREADME.md
Original
We have developed an unsupervised technique for MRI super-resolution. We leverage a recent pre-trained Brain LDM for building powerful image priors over T1w brain MRIs. Our method is capable of being adapted to different settings of MRI SR problems at test time. Our method try to find the optimal latent representation $z^∗$ in the latent space of the Brain LDM, which could be mapped to represent the SR MRI $G(z^∗)$.
This gif shows the image space of the gradual optimization process when InverseSR finds the optimal latent representation $z^∗$.
Install Requirements
pip install -r requirements.txt
Running InverseSR
We have given an example of ground truth high-resolution MRI ./inputs/ur_IXI022-Guys-0701-T1.nii.gz. The code of generating low-resolution MRI is contained. Please download the Brain LDM parameters ddpm and decoder from here into the InverseSR folder. Commands and parameters to run InverseSR can be found in job_script/InverseSR(ddim).sh and job_script/InverseSR(decoder).sh file.
Data Preparation
!! This model needs to be run on GPU/CPUs with at least 80GB of memory
You can find the necessary files for running the code here