Boosting Brain-to-Image Decoding with TRIBE v2 Data Augmentation
Researchers have developed a method to improve brain-to-image decoding by augmenting limited fMRI datasets with synthetic data. They utilized TRIBE v2, a large model trained on over 1000 hours of fMRI responses, to generate this synthetic data. Experiments on two datasets showed up to a 68% improvement in image retrieval accuracy compared to using only real data, demonstrating the potential for large-scale models to enhance data efficiency in brain decoding tasks. AI
IMPACT Enhances data efficiency for brain decoding tasks, potentially enabling new applications in neuroscience and AI.