Abstract: We introduce a modular, hybrid decoding architecture that separates motion tracking (“where”) and image reconstruction (“what”) to recover high-resolution visual stimuli from noisy retinal ...