🚀 Costin Chitic

Search

SearchSearch

Recent Notes

GoPro Livestream

May 13, 20252 min read

Mission: Adding timestamps to GoPro footage and using it in ROS for SeaClear.

Sources: This response on official community support, another community question, adding overlay, firmware update

Update Firmware tutorial: https://github.com/gopro/labs/tree/master/docs/install

Add an Overlay to Video

Here I can overlay HH:MM:SSaa on the real-time footage.

UTC Date And Time

PRO: There is a field in the metadata that could give the time format in ms.

Look into gopro-telemetry (Python) or gpmf-parser (C library).

CON: Only in postprocessing can I access this data.

UPDATE: I can only extract through gpmf-parser in offline, no online solution. Theoretically, I only need the .mp4 which has the metadata included.

Install gpmf-parser (C library)

Steps:

  1. git clone https://github.com/gopro/gpmf-parser.git
  2. cd gpmf-parser/
  3. mkdir build
  4. cd build
  5. cmake ..
  6. make

STMP and GPSU fields

From the official documentation, it appears that

  • STMP is a precise but relative microsecond timestamp — 1 Hz
  • GPSU is an absolute UTC time of low frequency that can miss here and there — 1 Hz

TO GET THESE FIELDS, ALWAYS GET THE VIDEOS DIRECTLY FROM THE SD-CARD OR CONNECT THE GOPRO THROUGH USB-C CABLE.

Metadata is lost if videos are sent through conventional platforms such as WhatsApp or AirDrop. Lost a lot of fucking time figuring out that one :)

To get them in code and actually view them, I added extract_utc.c(add link later) inside gpmf-parser/demo

STMP microsecond timestamp: 0.102411 s
GPSU UTC Timedata: 2015-04-25 10:38:23.679
STMP microsecond timestamp: 1.099822 s
GPSU UTC Timedata: 2015-04-25 10:38:26.319
STMP microsecond timestamp: 2.102296 s
GPSU UTC Timedata: 2015-04-25 10:38:27.309
STMP microsecond timestamp: 3.104771 s
GPSU UTC Timedata: 2015-04-25 10:38:28.299
STMP microsecond timestamp: 4.107248 s
GPSU UTC Timedata: 2015-04-25 10:38:29.344
STMP microsecond timestamp: 5.104663 s
GPSU UTC Timedata: 2015-04-25 10:38:30.334
STMP microsecond timestamp: 6.107141 s
GPSU UTC Timedata: 2015-04-25 10:38:31.324
STMP microsecond timestamp: 7.109622 s
GPSU UTC Timedata: 2015-04-25 10:38:32.314
STMP microsecond timestamp: 8.107037 s
...

So if STMP = 0.102411 s and GPSU = 2015-04-25 10:38:23.679, it means that:

  • The first STMP in the payload occurred 102 ms after the GPSU anchor.
  • The next STMP (~1.099 s) is ~1 second later → makes sense, it’s a ~30 Hz stream.

For some reason, there is always a drop in the GPSU time and then it stabilizes around ~1 second per frame.

CONCLUSION

I can't extract the timestamp attached to each frame since the sensors are working only at 1 Hz.

Graph View

Backlinks