Dataset for Cross-Modal Generalization of Neuro-Informed Generative Connectome (NIGC) in Visual Tasks

Overview: This dataset is provided as a supplementary resource for the "Neuro-Informed Generative Connectome (NIGC)" project, specifically supporting the cross-modal generalization experiments conducted on the mouse visual pathway using a video classification task.Data Components:microdata.zip (Required): Contains the 3D spatial coordinates of neurons for brain regions involved in the visual pathway. This data is sourced and preprocessed from the Blue Brain Project Cell Atlas and is essential for generating the anatomical structural priors.video classification_raw_data_mp4.zip (Raw Dataset): Contains the raw MP4 video files used for the 3-class video recognition task. These videos provide the dynamic visual stimuli required to drive the simulated retinal models.feature.zip (Pre-extracted Features / Cache): Contains the intermediate feature data processed by the Retina.py script. The script simulates retinal mechanisms to extract spatiotemporal features serving as inputs to the Superior Colliculus (SC) and Lateral Geniculate Nucleus (LGN). Because this extraction process is highly computationally expensive and time-consuming, this pre-computed cache is provided to allow researchers to bypass the preprocessing step and directly run the reservoir computing simulations (video_classification_reservoir.py).Usage: Please refer to the visual_task section in the official NIGC GitHub repository for detailed placement and execution instructions.