Multimodal deep neural decoding reveals highly resolved spatiotemporal profile of visual object representation in humans.