sebasmos/latent-sr-weights

Latent-SR: Domain-Specific Diffusion Weights for Medical Image Super-Resolution

Trained diffusion model checkpoints from the paper:

"Domain-Specific Latent Representations Improve the Fidelity of Diffusion-Based Medical Image Super-Resolution"
Sebastian Cajas, Ashaba Judith, Rahul Gorijavolu, Sahil Kapadia, Hillary Clinton Kasimbazi, Leo Kinyera, Emmanuel Paul Kwesiga, Sri Sri Jaithra Varma Manthena, Luis Filipe Nakayama, Ninsiima Doreen, Leo Anthony Celi.
arXiv:2604.12152 (2026) — under review at Nature Scientific Reports.

📄 Paper: https://arxiv.org/abs/2604.12152
💻 Code: https://github.com/sebasmos/latent-sr

Model Description

These are the trained x₀-prediction latent diffusion model (LDM) weights for 4× medical image super-resolution. Each checkpoint is the full LDM (UNet only; VAE is loaded separately at inference time).

All models were trained with:

• Schedule: cosine β, T=1000 train / T=100 inference
• Prediction target: x₀ (clean latent)
• Pipeline: frozen VAE encoder → diffusion UNet → frozen VAE decoder

Checkpoints

File	VAE	Dataset	SR PSNR (val)	Notes
medvae-4-3_brats.ckpt	MedVAE (3×64×64)	BraTS 2023 brain MRI	26.42 dB	Paper main result
medvae-4-3_cxr.ckpt	MedVAE (3×64×64)	MIMIC-CXR chest X-ray	28.87 dB	Paper main result
medvae-4-3_mrnet.ckpt	MedVAE (3×64×64)	MRNet knee MRI	25.26 dB	Paper main result
sdvae_brats.ckpt	SD-VAE (4×32×32)	BraTS 2023	23.51 dB	Baseline
sdvae_cxr.ckpt	SD-VAE (4×32×32)	MIMIC-CXR	25.58 dB	Baseline
sdvae_mrnet.ckpt	SD-VAE (4×32×32)	MRNet	22.34 dB	Baseline
medvae-4-1_brats.ckpt	MedVAE 4_1 (1×64×64)	BraTS	25.55 dB	Capacity control
medvae-4-1_cxr.ckpt	MedVAE 4_1 (1×64×64)	MIMIC-CXR	26.95 dB	Capacity control
medvae-4-1_mrnet.ckpt	MedVAE 4_1 (1×64×64)	MRNet	24.74 dB	Capacity control
medvae-8-1_brats.ckpt	MedVAE 8_1 (1×32×32)	BraTS	23.28 dB	Capacity control
medvae-8-1_cxr.ckpt	MedVAE 8_1 (1×32×32)	MIMIC-CXR	24.08 dB	Capacity control
medvae-8-1_mrnet.ckpt	MedVAE 8_1 (1×32×32)	MRNet	23.50 dB	Capacity control
medvae-8-4_brats.ckpt	MedVAE 8_4 (4×32×32)	BraTS	23.84 dB	Domain-matched control
medvae-8-4_cxr.ckpt	MedVAE 8_4 (4×32×32)	MIMIC-CXR	26.51 dB	Domain-matched control
medvae-8-4_mrnet.ckpt	MedVAE 8_4 (4×32×32)	MRNet	24.81 dB	Domain-matched control

Usage

from huggingface_hub import hf_hub_download
import torch

# Download the main MedVAE BraTS checkpoint
ckpt_path = hf_hub_download(
    repo_id="sebasmos/latent-sr-weights",
    filename="medvae-4-3_brats.ckpt"
)
state = torch.load(ckpt_path, map_location="cpu")

See the code repository for the full inference pipeline.

Citation

@article{cajas2026domain,
  title   = {Domain-Specific Latent Representations Improve the Fidelity of Diffusion-Based Medical Image Super-Resolution},
  author  = {{Sebastian Cajas} and {Ashaba Judith} and {Rahul Gorijavolu} and {Sahil Kapadia} and {Hillary Clinton Kasimbazi} and {Leo Kinyera} and {Emmanuel Paul Kwesiga} and {Sri Sri Jaithra Varma Manthena} and {Luis Filipe Nakayama} and {Ninsiima Doreen} and {Leo Anthony Celi}},
  journal = {arXiv preprint arXiv:2604.12152},
  year    = {2026},
  url     = {https://arxiv.org/abs/2604.12152}
}

Code: https://github.com/sebasmos/latent-sr · Paper: https://arxiv.org/abs/2604.12152

License

MIT License. The pretrained VAE weights (MedVAE, SD-VAE) are subject to their respective licenses.