BrainFLORA Dataset and Checkpoints

This repository hosts the released assets for BrainFLORA: Uncovering Brain Concept Representation via Multimodal Neural Embeddings.

• Paper: https://arxiv.org/abs/2507.09747
• Code: https://github.com/ncclab-sustech/BrainFLORA

BrainFLORA aligns EEG, MEG, and fMRI signals with visual-language representations for visual retrieval, image reconstruction, and image captioning.

Checkpoints

File	Purpose
checkpoints/eeg_01-06_01-46_150.pth	EEG single-modality retrieval encoder
checkpoints/meg_01-11_14-50_150.pth	MEG single-modality retrieval encoder
checkpoints/fmri_01-18_01-35_150.pth	fMRI single-modality retrieval encoder
checkpoints/Unified_EEG+MEG+fMRI_EEG_01-27_02-32_60.pth	Unified EEG+MEG+fMRI retrieval model
checkpoints/reconstruction_checkpoints/150.pth	Unified high-level reconstruction encoder
checkpoints/reconstruction_checkpoints/prior_diffusion/150.pth	Reconstruction diffusion prior
checkpoints/caption_checkpoints/90.pth	Caption-aligned unified encoder
checkpoints/caption_checkpoints/prior_diffusion/100.pth	Caption diffusion prior

External Caption Model

The Shikra model and mm_projector.bin are not hosted in this dataset repository. Download them from the original Shikra resources and place them under the code repository as:

external_models/shikra-7b
external_models/mm_projector.bin

References:

• Shikra paper: https://arxiv.org/abs/2306.15195
• Shikra code and model instructions: https://github.com/shikras/shikra
• Hugging Face model page: https://huggingface.co/shikras/shikra-7b-delta-v1

Usage

Clone or download this dataset repository, then follow the reproduction commands in the BrainFLORA code repository README.

git clone https://github.com/ncclab-sustech/BrainFLORA.git
cd BrainFLORA

Restore the downloaded assets so the code repository has the expected layout:

BrainFLORA/checkpoints/
BrainFLORA/features/
BrainFLORA/fmri_dataset/
BrainFLORA/meg_dataset/

Related Projects

BrainFLORA uses BrainHub-style caption metrics for caption evaluation. Please also cite the UMBRAE/BrainHub project when using these evaluation components.

• UMBRAE project: https://weihaox.github.io/UMBRAE/
• UMBRAE code: https://github.com/weihaox/UMBRAE
• BrainHub benchmark: https://github.com/weihaox/BrainHub

Citation

@inproceedings{li2025brainflora,
  author = {Li, Dongyang and Qin, Haoyang and Wu, Mingyang and Wei, Chen and Liu, Quanying},
  title = {BrainFLORA: Uncovering Brain Concept Representation via Multimodal Neural Embeddings},
  year = {2025},
  isbn = {9798400720352},
  publisher = {Association for Computing Machinery},
  address = {New York, NY, USA},
  url = {https://doi.org/10.1145/3746027.3754996},
  doi = {10.1145/3746027.3754996},
  booktitle = {Proceedings of the 33rd ACM International Conference on Multimedia},
  pages = {5577--5586}
}


@inproceedings{xia2024umbrae,
  author = {Xia, Weihao and de Charette, Raoul and Oztireli, Cengiz and Xue, Jing-Hao},
  title = {UMBRAE: Unified Multimodal Brain Decoding},
  booktitle = {European Conference on Computer Vision (ECCV)},
  year = {2024}
}