Image Generation using EEG data: A Contrastive Learning based Approach

In recent years, there has been active research on Neural Decoding aims to decrypt perceptual cognitive content, recall content, and motor information directly from brain signals. Simultaneously, deep learning has been widely implemented especially in generative models. Diverse architectures and learning methods have been formulated and applied across numerous fields. In this work, we focus on generating perceptual and cognitive contents using brain electroencephalography (EEG) data. We propose and demonstrate the efficacy of contrastive learning to generate images only from EEG data. We compare the performance of two electrode settings 1) visual cortex and 2) motor cortex.