Open-source code released by UMRAM research groups. Most repositories accompany published papers.
ICON Lab
Deep learning methods for medical imaging: MRI reconstruction, multi-modal image synthesis, diffusion and state-space models, and medical vision-language models.
- All repositories on GitHub
- SynDiff adversarial diffusion for image synthesis
- ResViT residual vision transformers for synthesis
- I2I-Mamba state-space image-to-image translation
- FDB Fourier-constrained diffusion bridges for MRI
- MambaRoll physics-driven MRI reconstruction
- MedTrim aligning medical vision-language models
Aykut Koç Lab
Machine learning and signal processing, extending into natural language processing, graph neural networks and the fractional Fourier transform.
- All repositories on GitHub
- Models and datasets on Hugging Face
- MTSIT multivariate time series imputation
- TrainableFrFT trainable fractional Fourier transform
- law-turk predicting Turkish court decisions
- scHyperLink gene regulation with hypergraph networks
- BitTransGNN quantized transformers with graph networks
Sarıtaş Lab
Methods for magnetic particle imaging and MRI, including correction of susceptibility artifacts in echo planar imaging.