Electronics, Год журнала: 2025, Номер 14(11), С. 2245 - 2245
Опубликована: Май 30, 2025
In contemporary visual decoding models, traditional neural network-based methods have made some advancements; however, their performance in addressing complex tasks remains constrained. This limitation is primarily due to the restrictions of local receptive fields and inability effectively capture information, resulting loss essential contextual details. Visual processing brain initiates retina, where information transmitted via optic nerve lateral geniculate nucleus (LGN) subsequently progresses along ventral pathway for layered processing. Unfortunately, this natural process not fully represented current models. paper, we propose a state-space-based model, SSM-VIDM, which enhances by aligning with brain’s mechanisms. approach overcomes limitations convolutional networks (CNNs) regarding fields, thereby preserving tasks. Experimental results demonstrate that model proposed study outperforms models terms exhibits higher accuracy image recognition Our research findings suggest based on pathway, can enhance performance.
Язык: Английский