MindSA

May 15, 2026 ยท View on GitHub

Official Implementation of MindSA in PyTorch

Brain2Text

  1. Download NSD

  2. Install StableDiffusionReconstruction

  3. Extract CLIP representations by running fmri_process/feature_extract_nsd.py

  4. Preprocess NSD by running fmri_process/NSD_handle.py

  5. Train the Reconstruction model by running brain2text_train_nsd.py

  6. Reconstructing text by running brain2text_infer_nsd.py

  • Our codebase builds on StableDiffusionReconstruction, MAE, SMALLCAP, stable-diffusion repositories. We would like to thank the authors.

Citation

@InProceedings{Chen_2025_ICCV,
    author    = {Chen, Jiaxuan and Qi, Yu and Wang, Yueming and Pan, Gang},
    title     = {Bridging the Gap between Brain and Machine in Interpreting Visual Semantics: Towards Self-adaptive Brain-to-Text Decoding},
    booktitle = {Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV)},
    month     = {October},
    year      = {2025},
    pages     = {21938-21948}
}