TLDR:
ffmpeg -y -i GOPR0001.MP4 -codec copy -map 0:m:handler_name:" GoPro MET" -f rawvideo GOPR0001.bin
Note the gap before GoPro MET should be a TAB, not a space.
-
gopro2json -i GOPR0001.bin -o GOPR0001.json
-
There is no step 3
I forked stilldavid's project ( https://github.com/stilldavid/gopro-utils ) to achieve 2 things:
- Export the data in csv format from /bin/gpmdinfo/gpmdinfo.go
- Allow the project to work with GoPro's h5 v2.00 firmware
This is my first Github repository. Any possible wrong practices are not intentional. Here continues Stilldavid's work: ##############################################################################################################
I spent some time trying to reverse-engineer the GoPro Metadata Format (GPMD or GPMDF) that is stored in GoPro Hero 5 cameras if GPS is enabled. This is what I found.
Part of this code is in production on Earthscape; for an example of what you can do with the extracted data, see this video.
If you enjoy working on this sort of thing, please see our careers page.
The metadata stream is stored in the .mp4
video file itself alongside the video and audio streams. We can use ffprobe
to find it:
[computar][100GOPRO] ➔ ffprobe GOPR0008.MP4
ffprobe version 3.2.4 Copyright (c) 2007-2017 the FFmpeg developers
[SNIP]
Stream #0:3(eng): Data: none (gpmd / 0x646D7067), 33 kb/s (default)
Metadata:
creation_time : 2016-11-22T23:42:41.000000Z
handler_name : GoPro MET
[SNIP]
We can identify it by the gpmd
in the tag string - in this case it's id 3. We can then use ffmpeg
to extract the metadata stream into a binary file for processing:
ffmpeg -y -i GOPR0001.MP4 -codec copy -map 0:3 -f rawvideo out-0001.bin
This leaves us with a binary file with the data.
- ~400 Hz 3-axis gyro readings
- ~200 Hz 3-axis accelerometer readings
- ~18 Hz GPS position (lat/lon/alt/spd)
- 1 Hz GPS timestamps
- 1 Hz GPS accuracy (cm) and fix (2d/3d)
- 1 Hz temperature of camera
Data starts with a label that describes the data following it. Values are all big endian, and floats are IEEE 754. Everything is packed to 4 bytes where applicable, padded with zeroes so it's 32-bit aligned.
- Labels - human readable types of proceeding data
- Type - single ascii character describing data
- Size - how big is the data type
- Count - how many values are we going to get
- Length = size * count
Labels include:
ACCL
- accelerometer reading x/y/zDEVC
- deviceDVID
- device ID, possibly hard-coded to 0x1DVNM
- devicde name, string "Camera"EMPT
- empty packetGPS5
- GPS data (lat, lon, alt, speed, 3d speed)GPSF
- GPS fix (none, 2d, 3d)GPSP
- GPS positional accuracy in cmGPSU
- GPS acquired timestamp; potentially different than "camera time"GYRO
- gryroscope reading x/y/zSCAL
- scale factor, a multiplier for subsequent dataSIUN
- SI units; strings (m/s², rad/s)STRM
- ¯\_(ツ)_/¯TMPC
- temperatureTSMP
- total number of samplesUNIT
- alternative units; strings (deg, m, m/s)
Types include:
c
- single charL
- unsigned longs
- signed shortS
- unsigned shortf
- 32 float
For implementation details, see reader.go
and other corresponding files in telemetry/
.