Repository navigation
Warn, rather than refuse, when a GPX track and a video differ in length - #944
Merged
Merged
Conversation
…apillary#940) This reverts the merge of mapillary#940 (3fcf6ca), back to main as it was after mapillary#939. mapillary#940 placed a GPX track along a video without GPS timestamps by the video's creation time. It was decided, after mapillary#940 was opened, that the creation time, which comes from the camera's clock, is never to be used to place a track: only a video's own GPS timestamps and its duration are. mapillary#944 handles videos without GPS timestamps that way instead, warning when the video and the track differ in length, and is stacked on mapillary#939, so it applies once mapillary#940 is reverted.
A video without a GPS clock of its own gives no trustworthy time to place a GPX track by: its creation time and file times come from the camera's clock. The track, whose times come from GPS, is then assumed to start with the video, in process and in video_process alike. When the track lasts more than a minute longer or shorter than the video, geotagging warns that the video start time needs aligning with the track, with what to do about it. A longer track warns only if it also lasts more than 1.5 times as long as the video, and the warning mentions timelapses, whose tracks last longer by design. Nothing is refused any more: sampling used to refuse a track that ran on past twice the video, and now places it from its start as process does. With --interpolation_offset_time the offset places the video along the track, and nothing is warned of. The warning is a MapillaryGPXAlignmentWarning, kept by the extractor with the durations and times of the track and the video, for a later change to pass on to the process report. Interval sampling stamps the samples of such a video from the start of the GPX track that geotags them, rather than from its creation time, which it now only falls back to for a video with neither a GPS clock nor a track.
caglarpir
force-pushed
the
align-gpx-with-video-start
branch
from
October 30, 2026 09:28
0505300 to
6b2095d
Compare
caglarpir
added a commit
to caglarpir/mapillary_tools
that referenced
this pull request
Oct 30, 2026
- A track assumed to start with a video without GPS is never refused for its length, in process or in sampling, and the creation time is never used to place it: a MapillaryGPXAlignmentWarning says when the track and the video differ in length (mapillary#944). Only --interpolation_use_gpx_start_time still refuses a long track. - A file that a GPX or NMEA track missed fails with the miss when no later source geotags it (mapillary#938). - The process report lists the geotagging warnings (mapillary#945). - upload caches the checksums of the files, and shows progress while computing them (mapillary#942, mapillary#943).
caglarpir
added a commit
to caglarpir/mapillary_tools
that referenced
this pull request
Oct 30, 2026
A GPX is assumed to start with a video without GPS, in process and in every kind of sampling, with a warning when the two differ in length, and is never placed by the creation time: only --interpolation_use_gpx_start_time still refuses a long track, in video_process. The process report lists the geotagging warnings.
caglarpir
added a commit
that referenced
this pull request
Nov 3, 2026
A video without a GPS clock of its own gives nothing to place a GPX track by, and the track is assumed to start with it. A track that runs on past the end of the video for longer than the whole video was not recorded for the video alone: the video could start anywhere along it. Sampling refused such a track from #841 until #944, which only warns that the video start time needs aligning with the track. Since then, sampling places the video at the start of whatever the track recorded, such as the first trip of a log of several. Sampling refuses it again, by distance and by interval alike, with the MapillaryGPXVideoStartUnknownError and the message it gave before, which for a track of a day or more says that the track covers more than the video. Nothing is refused when --video_start_time places the video along the track, when --interpolation_offset_time looks the video up along it, or when the video spans the track at --video_duration_ratio. A track shorter than the video, or longer by less than the video lasts, keeps the warning. process, which takes neither --video_start_time nor --video_duration_ratio, still uploads such a track whole, with the warning, as the track of a timelapse runs on past the end of its video too. It refuses only a track of a day or more, a log of several trips whose start is that of the first, unless the video lasts more than half as long. The creation time of the video still places nothing.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Stacked on #946, which is to be merged first; review the last commit.
A video without a GPS clock of its own gives no trustworthy time to place a GPX track by: its creation time and file times come from the camera's clock, which can be off, and mark the start or the end of the recording depending on the camera. The GPX, whose times come from GPS, is taken to be right. Without
--video_start_time, the track is now assumed to start with the video inprocessand in every kind of sampling, and is never refused for its length.When the track lasts more than a minute longer or shorter than the video, geotagging warns:
A longer track warns only if it also lasts more than 1.5 times as long as the video, and only then does the warning mention timelapses, whose tracks last longer by design. In sampling, the advice is
--video_start_timeand, for a timelapse,--video_duration_ratio. With--interpolation_offset_time, the offset places the video along the track, and nothing is warned of.--video_start_timeprocess, track more than twice the videoprocess, track 1.5 to 2 times the video, and more than a minute longerprocess, track more than a minute shorterprocess,--interpolation_offset_time, track more than twice the videovideo_processby distance, or with--video_geotag_source gpx, track more than twice the videoMapillaryGPXVideoStartUnknownErrorvideo_processby distance, track 2 s to twice the video longervideo_processby interval,--geotag_source gpx--interpolation_use_gpx_start_timestill refuses, in sampling, a track too long to end with the video at--video_duration_ratio. A video whose GPS has no timestamps is still sampled by interval with its creation time as the capture time, when no GPX geotags the samples: its positions come from its own GPS, and the time only labels them.The warning is a
MapillaryGPXAlignmentWarning, which carries the durations and the start and end times of the track, for a follow-up to list in the process report. #946 reverts #940, which placed the track by the creation time: this handles such videos instead.Tests: unit and integration suites, with and without ExifTool.