Inter-individual deep image reconstruction via hierarchical neural code conversion

Jun Kai Ho, Tomoyasu Horikawa, Kei Majima, Fan Cheng, Yukiyasu Kamitani

Abstract: Highlights•Neural code converters, which are trained to predict brain activity patterns from one to another individual when presented with the same stimulus, automatically learn the hierarchical correspondence of visual areas.•Converted brain activity patterns can be decoded into hierarchical DNN features to reconstruct visual images, even though the converter is trained on a limited number of data samples.•The information of hierarchical and fine-scale visual features is preserved with the functional alignment to capture the richness of visual perception.