HKU-MedAI/GEM-3D

[IJCV'2026] Generative Enhancement for 3D Medical Images

46
/ 100
Emerging

Implements a two-stage latent diffusion framework with KL-VAE compression for 3D volumetric synthesis, supporting both full-volume generation and position-conditioned slice-by-slice reconstruction. Leverages nnUNet preprocessing and integrates pretrained LDM components, enabling flexible conditioning modes (image-to-image, guidance) across brain and abdominal datasets. Optimized for multi-GPU training (8× V100/A100) with inference requiring <20GB memory.

No Package No Dependents
Maintenance 13 / 25
Adoption 9 / 25
Maturity 16 / 25
Community 8 / 25

How are scores calculated?

Stars

75

Forks

5

Language

Python

License

GPL-3.0

Last pushed

Mar 12, 2026

Commits (30d)

0

Get this data via API

curl "https://pt-edge.onrender.com/api/v1/quality/diffusion/HKU-MedAI/GEM-3D"

Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.