So I decided to try my analog video Evoque after about a year of not flying it.  It had my old rates and I really didn't feel comfortable flying aggressive with the analog feed (don't know how I ever did before).

So I attached my RunCam 5 Orange and sent it.  The Orange flaked on a bunch of videos (fragments only saved) but on the very first one it recorded the entire flight.

When I brought it into Gyroflow (just to add some lens correction) it started analyzing, so I left it and went off to do something else.  It ran for like 10 minutes and I was expecting the normal results I get from Gyroflow, but this time I was amazed.  I guess GyroFlow is doing a lot more of the grunt work these days.

Anyway, this isn't really a flying video, rather just some impressive GyroFlow handi-work with absolutely no input from me, other than selecting the lens filter.

I feel like gyroflow is always a hit or miss for me. I'm curious if the Orange provided better data than most or something? I haven't quite figured out the variables to consistently stabilize the videos well. Super smooth! I'm curious why lol

Gyroflow is improving with each version and getting easier to use with presets so some cameras literally just drop in the raw video. Some cameras have issues though, which camera are you using? The RC5 is actually more difficult to stabilize (so I've heard) because it drops lots of frames which throws off the gyro sync, the workaround was to add tons of sync points (that is in the RC5 preset) but maybe Runcam has fixed that?
@quadflyer nice job!!

Thanks guys!

So this is from the RunCam 5 Orange, with v1.2.2 firmware. Exact same setup I was trying with GyroFlow last fall. This version creates a .csv file with the gyro data.

After this video (and the flakyness I observed) I decided to update the firmware to 2.0.3, which creates a .gcsv file for the gyro data. I have yet to test fly and shoot a video with this version of firmware.

I know GyroFlow has gone through at least 2 revisions since I tried it last fall with this camera, when the results weren't that stellar. This last attempt with the latest version was amazing.

I transferred the TPU camera mount to my Nazgul (with WalkSnail), and removed the throttle limit of 87 and reset it to 100.
Set the Orange camera for 1440p (4:3) 1920x1440 and flew some test flights.
The original video was a little bumpy but GyroFlow took care of that

Wow.  Another 1/2 dozen or so flights with the RunCam 5 Orange, iFlight Nazgul 5", and GyroFlow with great results.
(latest versions of RunCam 5 Orange firmware and GyroFlow).
This was an unexpected, and pleasant, surprise to close out this summer.  This combo just works now.

Previously, my experience with GyroFlow (and the Thumb, my GoPro 5 Hero, RunCam 5 Orange, and WalkSnail Avatar Pro) has been meh, maybe one out of 10 videos would stabilize decently.  I would use GyroFlow regularly, but just for lens correction and aspect ratio tweaks. Now, every video from the RC5O stabilizes like I expected it to, with no/minimal user input.

And, the Nazgul isn't really tuned to carry the RC5O, so it bobbles when turning hard or with quick increased throttle - this I can clearly see in the Avatar HD goggles.  But, after running the footage through GyroFlow all of that nonsense is gone. I'm liking 35-50% Lens Correction strength these days too. 100% just stretches the image too much at the sides IMO.

Looking good and smooth. Is that dancing dust spec on the sensor, just removed a similar one from my camera, but I guess for the RC5 its fixed in place? If your RC is hard mounted, you could take its gyro feed to stabilize the Walksnail footage, wonder how that would look. The WS quality will be lower and probably have more sync issues, but curious how they compare.

Yeah, that 'spec' is inside the camera. Not ambitious enough to disassemble it yet - lol.
Good idea on stabilizing the WS footage with the RC5O data ... will give that a try now ...

Works really well Mstc!  It didn't at first, until I remembered about the offset thing:

• Rough Gyro Offset: -12.0s
  
• FOV: 1.75
  
• Smoothness: 0.150
  
• Zoom Speed: 2.000s
  
• Lens Correction Strength: 35%
  
• Rendered at 2160x1214 60fps 35Mbps
  

I'll compare stabilized RC5O to WS next ....

Wow, 12 seconds is hell of an offset.

Gyroflow can be hit and miss. It is mainly the sync challenges with the gyro data versus visual data.

(05-Sep-2023, 03:04 AM)ph2t Wrote: Wow, 12 seconds is hell of an offset.

Gyroflow can be hit and miss.  It is mainly the sync challenges with the gyro data versus visual data.

When using a different device to log gyro, unless you synchronize and start recording at exactly the same time, you will have an offset. If it is two videos you can just see the takeoff time point and enter the difference into the offset. If you click the checkbox next to it, gyroflow will check both the + and - offset for you so you don't have to figure out which is which. If you are matching to say a blackbox log, just find the arm timepoint on the log and use that to get the time difference with the arming beep/prop spin up on your video.

@quadflyer The log seemed to stabilize the dvr nicely, so which camera has the wider FOV? You can still see the small shakes though, maybe your RC5 is on a TPU mount vs the WS which is hard mounted to the frame.

Ahh - I missed that. The camera is in a tpu mount.

The Orange definitely has a larger field of vertical view - especially in 1440p 4:3 mode. Looks like WS has them beat horizontally.

Great side by side comparison. Which WS camera is that, doesn't the V2 come with a native 4:3 sensor? I seem to recall in 4:3 mode the onboard recording is still 16:9 but with black bars? or maybe I'm remembering wrong as that would be silly.

For flight footage, I prefer the wider FOV, and you have to admit it is pretty amazing that WS delivers action cam quality footage to your goggles feed

It's the Walksnail V1 Micro camera, which is native 16:9.  The newer V2 Pro is native 4:3 (but not the newer one with 'starlight' vision).  

And yes, the firmware will allow us to select 4:3 and it just cuts off the sides with these V1 cameras.  Silly indeed.  I don't know what gets recorded on the air unit or goggles however - I just saw the black bars in the goggles without any increase in vertical FOV, interpreted that as a useless option and didn't spend any time with it.

I agree, the WS video feed and wider FOV is great, no doubt!

Previous video updated with 3840x2160x60, 70Mbps variable bit rate ...

That 'spot' in the RC5O video appears to be a fleck, just below the outer surface of the protective glass. I can see it with a loupe. I can see other tiny white(?) flecks as well but this one is bigger and kinda fleshy-orange colored. Can't scrape it off from the outside, and it's not on the bottom (inside). Watched a couple of videos on disassembling the inners as well as removing the outer glass. Tried removing the outer glass, with heat, didn't budge. Don't want to break it. Not disassembling it either - no thank you. RunCam's glass appears to have lots of contaminants in it. I remember a similar spot on one of my Thumb cams as well.

Wait - so the latest version of gyroflow (1.5.4) can stabilize video even without gyro data ….

Apparently it extracts motion data by analyzing the video.

It doesn’t seem to work if something is in frame all the time, like an arm, stand-off, OSD text, etc. Props in view is okay since they’re fluttering/flickering and are not the same in every frame.

Takes a long time but seems to work well. It’s kinda flakey while it’s running, and at first I thought it locked up my computer, but let it run.

Don’t overwrite your current install of gyroflow. Some things don’t work the same or at all - this must be an alpha release …