correct motion run time and output

[AOONIH]

Hi, I've noticed 2 things when running the motion corrections. First thing is that the recording object post correction has fewer channels. Is that deliberate? Secondly, the estimate motion step is taking much longer that the detect and localise step, 200 s vs 1500s. Also wondering if that's common

Thanks

[David-H-Chang]

@AOONIH Hey, I don’t find it surprising that there will be fewer channels. Consider the top channel: its recordings are often incomplete due to motion, and sometimes it captures footage that actually belongs to the second or third top channels. As a result, complete recordings will be virtually nonexistent in both the top and bottom channels.

[samuelgarcia]

@AOONIH : you can control this behavior with border_mode="remove_channels" or "force_extrapolate" See https://github.com/SpikeInterface/spikeinterface/blob/main/src/spikeinterface/sortingcomponents/motion/motion_interpolation.py#L289