Generative Adversarial Network Based Image Reconstruction using Brain Activity Profi
In recent decades, the human-computer interaction field can expose the vision intelligence of human despite gathering their opinions about the scene descriptions. As per the human vision system, while a visual stimulus presented, the human brain will capture and preserve the image patterns at the visual cortex. Decoding the human brain activity profiles through functional Magnetic Resonance Imaging (fMRI) has achieved more attention in cognitive science. Especially, in contemporary cognitive neuroscience domain, this process would much support the neural encoding and decoding part of visual intelligence. There are few research claims in the reconstruction process of extracting the original vision stimuli from the captioned image patterns of the visual cortex. A visual image is assumed to be denoted at multiple spatial scales in the visual cortex, which may serve to retain the visual sensitivity by means of fine-to-coarse patterns at a single visual field location (Miyawaki, 2008). Hence, the scope for this reconstruction process is most probable while extracting the multiple spatial scales-based information via fMRI voxel data. In this paper, the Deep Learning approach of variations of Generative Adversarial Network-based Image Reconstruction (GAN-IR) model is used to reconstruct the visual stimuli.
Keywords- Functional Magnetic Resonance Imaging (fMRI); Generative Adversarial Network (GAN); Discriminator;Generator.