Mismatch between get_video_timestamps and get_position_timestamps

[samuelbray32]

Found a nwb file conversion where the position timestamps came out reasonable, but the video file object's timestamps were incorrect due to strange ptp-calibration during the experiment.
e.g. the HWTimestamp in this dataframe corresponds with a time in the 1980's

file = "/stelmo/sc4712/rats/SC1001/raw/20230905/20230905_SC1001_02_r1.1.videoTimeStamps.cameraHWSync"
data_file = pd.DataFrame(
        readTrodesExtractedDataFile(file)["data"]
    )
data_file

• We have addressed translating these incorrect timestamps in get_position_timestamps
• get_video_timestamps doesn't have these same features and takes values directly as reported from .cameraHWSync

Should we try to achieve similar timestamp mapping in the video_timestamp function to resolve this or leave attributed to errors at time of data collection?

[edeno]

As I recall, the reason for this is that it is just a little trickier to get the correct timestamps for the whole video. It would be nice to have correct, but I'm not sure what the solution is.

[samuelbray32]

Ah I see. There's no guarantee that the video is fully contained within the recorded .rec file's time range.

One option is when the timestamps are obviously incorrect, to use the PosTimestamp value in .cameraHWSync to map to the trodestimestamps in the rec file (similar to the non-ptp case in position tracking). For values that don't map we could put a nan value

[edeno]

Yeah, that would be an okay solution.

[sharon-chiang]

Hi @samuelbray32, I was wondering if you might be able to still help with this issue?

[samuelbray32]

@sharon-chiang sorry for delay. Can you try pulling the odd_video_timestamps branch and converting it again?

[edeno]

@sharon-chiang did any of the fixes work? Should this issue be closed?