3D single-level cell segementation for whole mouse brain

We worked on initial exploration of the dataset. Implemented different spike detection methods. Currently, we are exploring the approaches to maximize the interclass difference while minimizing the intraclass difference in neural spikes. We were able to obtain promising results with learnable distance metrics via NN-based methods. The low risks of the project include inconsistent single-unit electrode channels across days, independent spatial drift across different channels, and not having a ground truth for the spike sources. Not having a ground truth can be a blocker in the future if we want to benchmark the proposed approach against the SOTA.