InspektorGadjet
Second Officer
Flight distance : 439915 ft
Spain
Offline
|
vdefilippo Posted at 12-4 09:59
Okay, so I went out today for another flight. Here's a sample on full auto. I can clearly see the pulsing. Once you're looking for it, it's painfully obvious.
By the way, I noticed another issue. On my flight today, I must have forgotton to stop recording when I shut off the drone. The last mp4 file is now corrupt. On my Mavic Air, if you forget to stop recording when you shut off the drone, it still saves the file properly.
Apparently this happened to other Mavics too in the past, and is not something DJI can fix.
Pavel answer in my post with a good explanation of what is causing this effect.
Link to my post: https://forum.dji.com/thread-203144-1-1.html
This is his message:
I believe what you are seeing is a known artifact of imperfect video coding caused by accumulated error in GOP sequence that becomes visible on next I frame (because the error vanishes suddenly). The "breathing" occurs in your video at slightly more than one second distance (take a loudly ticking clock to see for yourself) and this well correlates with the fact that I frames in this video occur every 30 frames (while the video is 25 fps). You can see frame types by running:
ffprobe.exe -select_streams v -show_frames -show_entries frame=pict_type -of csv DJI_0057_EV_LOCKED.MP4
(you will also see that there are only I and P frames and no B frames which is sad).
The video with waterfall that you gave a link to is 30 fps and the "breathing" there occurs exactly each second while the I frames distance is again 30 frames so this supports my guess.
Unfortunately, video encoding in low-power CPUs (Mavic Mini has Ambarella H22) is usually performed by certain hardware, not software, so I would not count on firmware fix in the future. |
|