CortexMAE
Pretraining and inference for fMRI masked autoencoders using flat maps, parcels, or volumes. Includes pretrained checkpoints and an API for extracting embeddings.
MedARC-AI/CortexMAECode
Foundation models
Structural MRI foundation model
Masked-autoencoder pretraining for structural MRI, with data loaders, task-specific probes, and submission containers for the FOMO26 challenge.
MedARC-AI/smri-fmBrainLM
A foundation model for learning representations from brain activity recordings.
vandijklab/BrainLMEvaluation & benchmarking
Brainmarks
An extensible evaluation package for fMRI encoders, with frozen-backbone probes, logistic regression, and support for adding models and datasets.
MedARC-AI/BrainmarksMultimodal & generative models
Algonauts 2025
Video, audio, and language features combined to predict brain responses to natural movies. Fourth place in the 2025 challenge.
MedARC-AI/algonauts2025MindEye2
Shared-subject models for fMRI-to-image retrieval and reconstruction, with adaptation to a new participant using as little as one hour of data.
MedARC-AI/MindEyeV2MindEye
Image reconstruction from fMRI recordings in the Natural Scenes Dataset.
MedARC-AI/fMRI-reconstruction-NSDDNA-Diffusion
Diffusion probabilistic models for generating regulatory DNA sequences.
pinellolab/DNA-Diffusion