YRGB Color Managed -> Input Color Space for DJI Air?

Has anyone had luck with the "DLOG" footage from a DJI Air2s when working in the “DaVinci YRGB Color Managed” workspace (not ACEScc). There is already one titled “DJI D-Gamut/D-Log” but it must be for a different profile / drone as when applied it crushes black, blows out whites, and add a purple hue to shadows (and then oversaturated everything else!).
Thanks
Nathan

PS - I can upload a raw sample of the Drone Footage if that helps.

FYI - Here is link to a 7sec sample - https://behome.dyndns.info/index.php/s/smkD2tW5aeJGi6P

I'm not familiar with the DJI devices but it sounds like you shot in D-Log but not D-Gamut?
Your footage is probably already rec.709 as colorspace and only the gamma is from DJI so you would need to put a Colorspace Transform node on the footage converting only D-Log to sRGB or gamma 2.4 depending on your workflow and leave the colorspace the same.

If you want your entire project managed in this mode you can select use separate colorspace and gamma in the project color settings.

Thanks shebbe. From my testing, I too think the footage is rec.709 / D-Log, but it is applying the Davinci D-Log part is causing the issue. Here are two pics showing the footage / settings / scopes in the different combinations of ColorSpace / Gamma:
- 709 / 709 : Looks flat like other log footage but all the data is unclipped
- 709 / D-Log : Crushed Blacks & Blown Out Whites, looks like the wrong Gamma curve is being applied

I've tried most combinations of Colour Space and Gamma in Resolve but none look correct.

Thanks
Nathan

I just tried all the DJI LUTs supplied with Resolve and the only one that is even remotely usable on your footage, IMO, is the first one (DJI_Phantom3_DLOG2sRGB_Improv).

But I'd just make it look nice by eye.

Thanks - FYI the LUT for the 2s from the DJI site is OK (better than the ones with shipped with Resolve).... however I'm trying to not use LUTS at all, but the Davinci Color Managed workspace instead for the technical grade.

Hm yes that doesn't look very correct.
I have the feeling some kind of double conversion is happening. Are you sure that is what the "raw" image looks like? It doesn't look very log to me compared to other D-Log footage I've seen.

Maybe switch back to unmanaged YRGB and test with a CST instead going directly to rec.709 not to the intermediate timeline space.

When you are working in Managed all the clips get an Input Color Space assigned based on available metadata. They auto convert from that to the chosen timeline space and only unknown files are set to chosen input space. Might want to check what the files are set to, just right-click them in your bins.

Maybe care to share a frame of the raw data unmanaged so I can investigate too.

The sample clip is cut directly from the original footage straight out of the drone (no transcoding at all - I used ffmpeg to extract the video track between time markers)..... but, there is something odd about this footage as I redid that same shot today and this one looks OK using the inbuilt Input Colour space for DJI D-Gamut / D-Log on this clip (well it still looks like a consumer oversaturated image but that is the "look" I guess.... at least nothing is blown out or crushed).

Why are the different?
- One difference between the two clips is the first one had a V-ND on it and the 2nd did not. I need to test that more to see if that is an issue or not but checking the SRT file, Clip 1 was 100ISO @ 1/60th with ND and Clip 2 was 100ISO @ 1/350 with no ND (when the pool is dead centre)
- I've checked the SRT file and both claim to be D-Log and neither were at the minimum shutter and max ISO

Here is the reshoot V2 Clip- https://behome.dyndns.info/index.php/s/t4KAkNbnPNETkQa

shebbe, I really appreciate another set of eyes on this!


Thanks
Nathan

AFAIK - with YRGB Color Managed, resolved will (in order)
- If RAW detect it and apply the correct Input (I've nothing to test)
- If MetaData exists (MXF etc) detect it and apply the correct Input (eg my FX6)
- Otherwise default to what is set in "Input Colour Space" (eg my Air2s)
- You can override the above by manually assigning the "Input Colour Space"

The DJI footage has no meta data on Color Space so I double check it is assigned by using either the default Input Colour Space or assigning it directly. I then used this method to check what it looks like using all sorts of combinations of color space and gamut.

Hey Nathan,

I get a totally valid result using D Gamut / D Log.
There has to be something else you are missing here. Did you check if you don't have any Group or Timeline nodes doing things that you don't see directly? Maybe try the setup in a separate new project to verify. If it still gives the same result maybe check another / latest DaVinci Resolve version.

2021-07-08 11_06_58-2.jpg (274.83 KiB) Viewed 5264 times

Hey shebbe,
Yup, perfect! - that is also what I get with todays clips (V2). Looks valid as you say . The original one does not, so I'll need to see if it is the ND filter (or something else). If you now put both V1 and V2 on the same timeline see what you think of V1 vs V2!

Thanks again,
Nathan

In summary, here are the two clips (shot on different days):
- Clip 1: 100ISO @ 1/60th with ND Filter : https://behome.dyndns.info/index.php/s/smkD2tW5aeJGi6P
- Clip 2: 100ISO @ 1/350th with No Filer : https://behome.dyndns.info/index.php/s/t4KAkNbnPNETkQa

I now see what you mean.
This has nothing to do with the conversion but with the capture.
Looks like you are saturating the sensor with the first clip. Despite shooting log that detail is already gone in the capture.

What I recommend is trying to shoot at the camera's native ISO or even higher. That should preserve better detail in the highlights but you'll need a stronger ND to stay at 1/60th or shoot at a higher shutter.

The main difference as to why the second take looks better is because the sun is less present there so the scene has less dynamic range.
You can easily see that you miss just a slight amount of stops dynamic range when you match it to take 2.

3.jpg (442.31 KiB) Viewed 5250 times

Have a look at this video, he explains exposing a scene vs ISO in good detail.

It's interesting, the DJI 2s was set to allow both ISO and Shutter speed to change at will. I'd put the ND filter on the DJI 2s in Clip 1 to allow it to shoot at a lower shutter speed. According to the meta data, both clips were recorded using 100ISO (base for the DJI 2s) just the shutter speed was lower with the filter (as you would expect). The first clip with the ND filter had a lower shutter speed of 1/60 vs 1/350 for the 2nd clip (that did not have a ND Filter).... so the ND filter was doing it's job..... but the first clip has blown highlights and crushed blacks, but you can get the details back in the blacks by changing the colour space from DLOG to 709 (or Bypass) and pushing up the shadows - see pic (so the info is there).... but not the highlights.

Bypass or 709.jpg (962.26 KiB) Viewed 5234 times

I agree it is a capture issue - I'll really need to test the ND filters to see if they are throwing off the sensor in the Air 2S. The whole idea of the ND filters is to keep the shutter speed low for motion blur, but if it causes issues with exposure I can always add blur in post.

An ND doesn't change the DR, but more motion blur might influence the compression in the shadows.

An ND doesn't change the DR, but more motion blur might influence the compression in the shadows.

This.
So yea the shadows in my example were a bit messed up. Probably due to the conversion to DWG/Intermediate. You probably can make cleaner results if you don't use managed but instead have group or timeline nodes at the end of your grades converting your source to your output format. That way you still have manual control over what you adjust before which conversion.

But it's best to do a test and shoot at ISO 100/400/800/1600 and use your ND / shutter to compensate. Then you'll see the difference it makes in the bias for highlights and shadows.
Make sure the lighting condition of your scene doesn't change between takes and has roughly the range of your first clip.

FYI - I'm working with:

Tobias Wiedmer, CSI
Digital Colourist / Colour Scientist
www.salamifish.com

on this and taking some test footage to compare with the FX6 under the same conditions. I'll update on the results.

The exercise Tobias wanted me to do (take some control shots at the same time on both the DJI Air 2s and the Sony FX6) has already helped. While I've not yet heard back from him (it's the middle of the night for him), it appears that I can get a pretty good match by setting the Air 2s footage to 709/DGamut using the DaVinci settings below. Looks like you may be correct in that the footage is not LOG (even though that is what it is advertised as having and what the apps option is called). These same settings also work well for my earlier clips as well.

FX6 and Air 2s Matching.jpg (87.25 KiB) Viewed 5079 times