This is a reminder about where we left of in June 2024, thinking about integrating KIlosort 4 into the lab pipelines.
I did some reviewing and testing of Kilosort 4 with notes at pipelines/docs/kilosort-4.
It was a good experience! The Kilosort 4 seems much better thought-out from an open source source project. It was fun to work with. As a developer, this was a very different experience compared to working with Kilosort 3.
We could think about trying to replace just Kilosort 3 witih Kilosort 4. That's probably the shortest path to 4.
The Kilosort paper and docs also mention SpikeInterface a lot. We could probably do a bit more work and think about how to replace Kilosort 3 with a SpikeInterface step, rather than Kilosort 4 specifically. We'd need to think about the inputs and outputs to that step and possibly make a few format changes, but I think it's a realistic vision!
This is a reminder about where we left of in June 2024, thinking about integrating KIlosort 4 into the lab pipelines.
I did some reviewing and testing of Kilosort 4 with notes at pipelines/docs/kilosort-4.
It was a good experience! The Kilosort 4 seems much better thought-out from an open source source project. It was fun to work with. As a developer, this was a very different experience compared to working with Kilosort 3.
We could think about trying to replace just Kilosort 3 witih Kilosort 4. That's probably the shortest path to 4.
The Kilosort paper and docs also mention SpikeInterface a lot. We could probably do a bit more work and think about how to replace Kilosort 3 with a SpikeInterface step, rather than Kilosort 4 specifically. We'd need to think about the inputs and outputs to that step and possibly make a few format changes, but I think it's a realistic vision!