Watch, debate! Fake or real???

Interesting

Intersting to see   

I don’t think it’s interesting, more devastating if it’s true. But
Then again I can do the same test and take pictures of a cellphone, gopro etc and the other thing the mp2 is not in d-log!!

FOV Wide uses the whole 5k sensor, but scales it down to 4k video. This is why it looks like it does above.
If you want perfect video, use the other 4k setting which crops the 5k image to 4k video and does not suffer the problems above.
You lose some FOV but gain video quality, and an advantage in that you don't have to get so close.

Still images always use the whole sensor. If you set it for max still image resolution, and cropped 4k video, you will see the FOV change when you switch between photo/video mode.

Very interesting

mnix Posted at 2018-9-15 00:53
FOV Wide uses the whole 5k sensor, but scales it down to 4k video. This is why it looks like it does above.
If you want perfect video, use the other 4k setting which crops the 5k image to 4k video and does not suffer the problems above.
You lose some FOV but gain video quality, and an advantage in that you don't have to get so close.

Doesn’t really have anything to do with the use of the sensor or not. If it don’t read whole sensor in video we are all bying a lie. Not ok if true, but I assume this is bs!!!! The video that is

It is pretty congruent. This is actually common on budget cameras.

I'm not sure if line skipping can be done in an irregular pattern. The most intuitive approach is to skip even or odd lines. That would leave us with around 2.7K resolution, with aliasing, upscaled to 4K.

0Kajuna0 Posted at 2018-9-15 04:19
It is pretty congruent. This is actually common on budget cameras.

I'm not sure if line skipping can be done in an irregular pattern. The most intuitive approach is to skip even or odd lines. That would leave us with around 2.7K resolution, with aliasing, upscaled to 4K.

I don't think it is line skiping due to heat problem, skiping lines still needs a fast clock, the heat generated if lines are skipped is more less the same. I would rather suspect that they don't have enough processing power to do anti-aliasing filtering due to down-sampling at an awkward conversion ratio from 5260 to 4000 pixels. Those morie circles look to me are aliasing distortion. Lack of processing power is also the reason why they didn't do len distortion correction at 4k FOV mode, and possibly 4k 60fps mode.

Nice to see thorough testing on this particular aspect. Definitely sad to be disappointed by the performance in this respect. I was definitely hoping it was closer to the P4P and I could replace that with something more portable for the same work. Was afraid something like this was happening after starting to see some of the other less thorough comparison videos out there. At least some people will get to enjoy a bump in line from my cancelled pre-orders

fans799afcf3 Posted at 2018-9-15 06:54
I don't think it is line skiping due to heat problem, skiping lines still needs a fast clock, the heat generated if lines are skipped is more less the same. I would rather suspect that they don't have enough processing power to do anti-aliasing filtering due to down-sampling at an awkward conversion ratio from 5260 to 4000 pixels. Those morie circles look to me are aliasing distortion. Lack of processing power is also the reason why they didn't do len distortion correction at 4k FOV mode, and possibly 4k 60fps mode.

Well, yes. I don't know the reason they skip lines. It might be the overheating or it might be the processing power (which if you think about it, is related. You can demand more from a processor but it will get hotter). But the aliasing happens precisely because lines (or pixels) are being skipped.

TheNorthern Posted at 2018-9-15 02:41
Doesn’t really have anything to do with the use of the sensor or not. If it don’t read whole sensor in video we are all bying a lie. Not ok if true, but I assume this is bs!!!! The video that is

Video can be verified or debunked by repeating Chart test, if one has access to Phantom 4 Pro and Mavic 2 Pro.
If you have Mavic 2 Pro, you can at least confirm whether you see same video results with Chart.  Although getting distance drone was from chart same is going to be difficult.

Be interesting to see if video results for Mavic 2 Pro differ when videoing 30 fps vs. 15 fps.  To see if claimed line skipping is discontinued with fps cut in half.

It's strange that Totten states that he still absolutely loves the M2P.  Having generally inferior video capability compared to the Mavic 2 Zoom and the Mavic Air is a considerable shortcoming.  As I understand the situation, the problem with the resolution of the M2P only applies to video.  As a brand new owner of an M2P, at least I can take some comfort in the hope that I can get some decent still shots from interesting perspectives using the drone.  But I am very disappointed that DJI apparently did not apply their success with the Phantom 4 Pro's use of a 1" sensor to their use of the 1" sensor in the M2P.  Totten didn't mention the possibility of DJI coming up with a fix in a firmware update, so I assume that will not be possible.

Seems like in the comments on youtube that the sample was not in 10bit. In theory that would make a big deal?

TheNorthern Posted at 2018-9-15 09:08
Seems like in the comments on youtube that the sample was not in 10bit. In theory that would make a big deal?

The bits allow for extra shades (gradients) between no Red and all Red.  Same for Green & Blue.  
0,0,0 would be solid black.  1023,1023,1023 would be solid white.  As would 8-bit.  0,0,0 would be solid black and 255,255,255 solid white.  Being test chart is black lines on white, doubtful that extra bits would make differences as to differentiating lines.

HedgeTrimmer Posted at 2018-9-15 10:18
The bits allow for extra shades (gradients) between no Red and all Red.  Same for Green & Blue.  
0,0,0 would be solid black.  1024,1024,1024 would be solid white.  As would 8-bit.  0,0,0 would be solid black and 256,256,256 solid white.  Being test chart is black lines on white, doubtful that extra bits would make differences as to differentiating lines.

Cheers!!

But I can’t understand any of this. If I watch on my 4k monitor the MP2 images from 4k is superior to my old drones, mavic pro, mavic air, p3 and even inspire. Don’t get it.

TheNorthern Posted at 2018-9-15 10:22
Cheers!!

But I can’t understand any of this. If I watch on my 4k monitor the MP2 images from 4k is superior to my old drones, mavic pro, mavic air, p3 and even inspire. Don’t get it.

Understandable, I made a mistake.  I was fixing it as you were responding.  I put 256, instead of 255, same with 1024, instead of 1023.  There is 256 possible shades (0 to 255).

TheNorthern Posted at 2018-9-15 10:22
Cheers!!

But I can’t understand any of this. If I watch on my 4k monitor the MP2 images from 4k is superior to my old drones, mavic pro, mavic air, p3 and even inspire. Don’t get it.

Images as in snapshots or images as in video?
Perhaps difference in Lens used on older stuff vs. MP2?   Help to compare to MP2 to Phantom-4 Pro as was done in video.

HedgeTrimmer Posted at 2018-9-15 10:31
Images as in snapshots or images as in video?
Perhaps difference in Lens used on older stuff vs. MP2?   Help to compare to MP2 to Phantom-4 Pro as was done in video.

Video,  photo on air not so good ;)

HedgeTrimmer Posted at 2018-9-15 10:18
The bits allow for extra shades (gradients) between no Red and all Red.  Same for Green & Blue.  
0,0,0 would be solid black.  1023,1023,1023 would be solid white.  As would 8-bit.  0,0,0 would be solid black and 255,255,255 solid white.  Being test chart is black lines on white, doubtful that extra bits would make differences as to differentiating lines.

I'm not so sure it's that simple. The normal mode is processed (for color and the lens bend). The 10 bit dlog would be nice to see in this test. I don't prefer the normal mode on the m2. It's like dji spiked the color wheel in certain spots.

Lemme clip a video here of auto shutter dloghq off my first flight (graded, no lut, sun out of control). I like the dloghq from what I've seen with no nd. Big step up for me. M1 zoomed in your looking at donkey kong. Interesting to see!

thehippoz Posted at 2018-9-15 11:54
I'm not so sure it's that simple. The normal mode is processed (for color and the lens bend). The 10 bit dlog would be nice to see in this test. I don't prefer the normal mode on the m2. It's like dji spiked the color wheel in certain spots.

Lemme clip a video here of auto shutter dloghq off my first flight (graded, no lut, sun out of control). I like the dloghq from what I've seen with no nd. Big step up for me. M1 zoomed in your looking at donkey kong. Interesting to see!

It's like dji spiked the color wheel in certain spots.

Quite possible.  I belive that is something done by all camera makers, with aid of maker of sensor used in camera.  Explanation being sensors do not respond linearly across light spectrum and darkness to brightness levels.  Along with camera makers trying to better their competitors, by little tweaks.


Still feel, the possibility demonstrated in video, needs independent verification.  There has to be others out there that have Phantom 4 Pro and Mavic 2 Pro they can test with using a Resolution Chart.  


Having already seen a video whereby the MP2 appears to beat Phamtom in resolution in dim lighting.  The verification of video's test needs to be done under bright lighting, regular room lighting, and dim lighting.

HedgeTrimmer Posted at 2018-9-15 13:42
It's like dji spiked the color wheel in certain spots.

Quite possible.  I belive that is something done by all camera makers, with aid of maker of sensor used in camera.  Explanation being sensors do not respond linearly across light spectrum and darkness to brightness levels.  Along with camera makers trying to better their competitors, by little tweaks.

Yeah I'm sure someone will do it. Guess could print up one of those charts and see what happens. I have the pro. Damn I need to change the link of that video I posted too. What a mess youtube is, if you clip a video from another video- it makes a 1080p flv of it

Bah maybe this guy is a prankster. I just ran some tests and took screens, played back in vlc on a 4k monitor. These are tiff uncompressed. Look at the hq.

https://www.dropbox.com/s/voasvgis3mn43jv/HP.7z?dl=1

Source http://www.bobatkins.com/photogr ... lens_sharpness.html

M2P  has a resolution falling when the F value is 5.6 or more.
ZOOM looks sharp but watercolor occurs as mavic pro.
It is necessary to evaluate the watercolor resolution with the F value and sharpness setting as parameters.

Copy paste from a Facebook discussion, permission given to copy it:

John Phillips: From someone who works in TV for the past 40 years... What this guy is claiming is totally bogus.  1080 and 4K are exactly the same resolution.  4K is just 4x 1080 pictures joined together to give you more real estate.  The only real benefit of 4K is in editing you have a lot more latitude to digitally crop a piece of the picture and still have 1080 resolution.  That's it.  

Where this guy is incorrect is that the resolution changes depending on what mode you're in.  In fact is, in either mode you still have either 4K (3840x2160) or HD (1920x1080) resolution being recorded.  Because they're all fixed focal length lenses, what changes is the processing that zooms them in to change the field of view.  It does not change the recorded resolution but like any digital zooming it will slightly degrade the actual picture quality.  

Resolution in television was originally measured in scan lines back in the days of SD but now it's measured in pixel lines in digital.  So no matter what mode you're in, you're still getting 2160 (4k) or 1080 (HD) lines of resolution.  

When he talks about "pixel skipping," what he's really referring to is compression, not changes in resolution.  Video is compressed to save data in the recording.  More compression = poorer picture.  

As for his claim of "line skipping," I'd like to see some proof of his claim that they're skipping scan lines on the sensor.  If this were true, it would mean that you're not getting a true HD image out of the sensor and that makes absolutely no sense.  If you eliminate half the lines of resolution you're reducing the image down to slightly higher than SD resolution.  That would be extremely noticeable and most people would find it unacceptable in any conditions.  

As for his tests, If this guy was doing a legitimate test of the images he would have had a serious set of SMPTE camera charts to show both resolution and coloremetry.  The fact that he printed out a chart and it's not correctly scaled tells me he has no clue what he's talking about.  Firstly what he used was not a legitimate SMPTE resolution chart.  The link below shows what legitimate SMPTE (Society of Motion Picture and Television Engineers) resolution charts look like.  They start at about $2 grand each.  Clearly not something you print off of the internet.  Besides which, there is no home printer capable of printing the kind of resolution required for a real resolution chart to actually work.  http://dsclabs.com/test-charts/resolution-focus-test-charts/

The last thing that blew all credibility for me of his claims is that he was using some kind of pot with a piece of the field turf from the super bowl in it as a reference for focus.  What a joke.   This is no legitimate test.  Again there are proper, calibrated focusing charts available to do this properly.  

So in short, I think you can distill the changes in the images down to two things... video compression and processing and likely quality of the glass in front of the sensor.  Let's face it, folks, you're not getting the same quality glass in a $1200 drone as you would in a $150k UHD television camera lens.  

FYI, this is what a $150k television camera lens looks like... and there's a reason why they cost $150k:  https://www.usa.canon.com/internet/portal/us/home/products/details/lenses/broadcast/4k-uhd-field-box-lenses/uhd-digisuper-122

The only one thing he said about the drones that made any sense is that they all look pretty good.

fans799afcf3 Posted at 2018-9-15 06:54
I don't think it is line skiping due to heat problem, skiping lines still needs a fast clock, the heat generated if lines are skipped is more less the same. I would rather suspect that they don't have enough processing power to do anti-aliasing filtering due to down-sampling at an awkward conversion ratio from 5260 to 4000 pixels. Those morie circles look to me are aliasing distortion. Lack of processing power is also the reason why they didn't do len distortion correction at 4k FOV mode, and possibly 4k 60fps mode.

That's exactly what I'm thinking also
Remember, DJI says they've not enough power to fix the lens distorsion in Full FOV. Then...

castormalin Posted at 2018-9-16 00:24
That's exactly what I'm thinking also
Remember, DJI says they've not enough power to fix the lens distorsion in Full HD. Then...

They have officially stated that??

TheNorthern Posted at 2018-9-15 23:57
Copy paste from a Facebook discussion, permission given to copy it:

John Phillips: From someone who works in TV for the past 40 years... What this guy is claiming is totally bogus.  1080 and 4K are exactly the same resolution.  4K is just 4x 1080 pictures joined together to give you more real estate.  The only real benefit of 4K is in editing you have a lot more latitude to digitally crop a piece of the picture and still have 1080 resolution.  That's it.  

As he says, the test protocol is not perfect but the point is not to make absolute measurements, the point is to make a comparison.
How Mr '40 years of TV work' explains the difference ?

TheNorthern Posted at 2018-9-16 00:32
They have officially stated that??

It was mentioned in one of the threads related to the lens distorsion by one DJI Tech. Actually, it's a kind of choice between 10 bits and distorsion correction. 10bits requires more power.

All 4k30, didn't move anything between hq and fov either. Uploaded in h265. Maybe a little while to process for youtube but it should be in 4k

thehippoz Posted at 2018-9-16 02:30
https://www.youtube.com/watch?v=Mm2F9NZ_o4M

All 4k30, didn't move anything between hq and fov either. Uploaded in h265. Maybe a little while to process for youtube but it should be in 4k

Wow there is a MASSIVE difference between HQ and the rest!!! It’s sharpness i massive. Odd since the youtuber claims therr is no difference in d-log hq 10bit!! Thanks

TheNorthern Posted at 2018-9-16 06:21
Wow there is a MASSIVE difference between HQ and the rest!!! It’s sharpness i massive. Odd since the youtuber claims therr is no difference in d-log hq 10bit!! Thanks

Yeah np! I was watching twitch last night and it bothered me. The hq is 1:1 crop of the sensor, and ends up being a nice zoom too.

Imo hlog fov is better for night time because there's no mosquito noise. That Hq is clean.

TheNorthern Posted at 2018-9-15 23:57
Copy paste from a Facebook discussion, permission given to copy it:

John Phillips: From someone who works in TV for the past 40 years... What this guy is claiming is totally bogus.  1080 and 4K are exactly the same resolution.  4K is just 4x 1080 pictures joined together to give you more real estate.  The only real benefit of 4K is in editing you have a lot more latitude to digitally crop a piece of the picture and still have 1080 resolution.  That's it.  

The fact that he printed out a chart and it's not correctly scaled tells me he has no clue what he's talking about.

Think Facebook argument is a Red Herring in way printed chart was used in video.  Remember he was comparing drones cameras against each imaging same printed chart.

Argument would be valid if he were only evaluating MP2 camera using printed chart, against camera images taken against purchased official chart.

Still, the video test done needs to be independently verified or debunked.

fans799afcf3 Posted at 2018-9-15 06:54
I don't think it is line skiping due to heat problem, skiping lines still needs a fast clock, the heat generated if lines are skipped is more less the same. I would rather suspect that they don't have enough processing power to do anti-aliasing filtering due to down-sampling at an awkward conversion ratio from 5260 to 4000 pixels. Those morie circles look to me are aliasing distortion. Lack of processing power is also the reason why they didn't do len distortion correction at 4k FOV mode, and possibly 4k 60fps mode.

Wonder about that too.  Skipping lines, would only save energy in processing of lines skipped.  The sensor chip would need a Energy Saving (resulting less heat) feature/option built-in.

Would be nice to compare the same image of a resolution chard in photo mode (DNG or even jpeg) and in 4K full FOV mode.
The compression does not degrade a lot Black&White synthetic images, and quasi not at all regarding resolution.
I mean, 1/ without moving the distance sensor - chard and 2/ with the same image width

castormalin Posted at 2018-9-16 09:36
Would be nice to compare the same image of a resolution chard in photo mode (DNG or even jpeg) and in 4K full FOV mode.
The compression does not degrade a lot Black&White synthetic images, and quasi not at all regarding resolution.
I mean, 1/ without moving the distance sensor - chard and 2/ with the same image width

Hard to tell from the guys video, but I thought when he was saying ''zooming'' in, he was doing zooming after the fact.  Zooming was being done on playback, so it would easier for viewers to see circles.

If he did "zooming" in, using digital "zoom" of drone's cameras, then his video is useless.

Really need more details as to what he did.  He does say he used playing cards to frame the edges.  And he does acknowledge use of printed chart and why it is okay in this case.

castormalin Posted at 2018-9-16 09:36
Would be nice to compare the same image of a resolution chard in photo mode (DNG or even jpeg) and in 4K full FOV mode.
The compression does not degrade a lot Black&White synthetic images, and quasi not at all regarding resolution.
I mean, 1/ without moving the distance sensor - chard and 2/ with the same image width

Well thing is, it takes over an hour on my rig to render 30 seconds of h265 at 100mbps cbr high profile. That's 100% on all 6 cores xd Only way to get decent 10 bit youtube videos. Else it's 264 all day.

You guys can do it if you want, but think my video shows what the m2 is and isn't. The guy in op video is classic youtube. Like the guy flying around in 8 bit showing night compares against other 1/2" sensor drones

thehippoz Posted at 2018-9-16 09:58
Well thing is, it takes over an hour on my rig to render 30 seconds of h265 at 100mbps cbr high profile. That's 100% on all 6 cores xd Only way to get decent 10 bit youtube videos. Else it's 264 all day.

You guys can do it if you want, but think my video shows what the m2 is and isn't. The guy in op video is classic youtube. Like the guy flying around in 8 bit showing night compares against other 1/2" sensor drones

Well thing is, it takes over an hour on my rig to render 30 seconds of h265 at 100mbps cbr high profile. That's 100% on all 6 cores

Nothing against your rig.  I can see many people running into that issue.  
Really hard to justify spending big bucks on a personal rendering farm, when it will sit idle most of time.

With that said, has anyone looked into purchasing time on a rendering farm (aka Rent-a-Render - cloud computing)?

HedgeTrimmer Posted at 2018-9-16 10:27
Well thing is, it takes over an hour on my rig to render 30 seconds of h265 at 100mbps cbr high profile. That's 100% on all 6 cores

Nothing against your rig.  I can see many people running into that issue.  

I have a linux rig that runs pyrit Cloud computing be fun. I don't have any real reason to upload in h265. If I did, I'd invest in a 10 core. The chip alone for x99 is over 800 right now. I don't think could justify that kind of money.

thehippoz Posted at 2018-9-16 10:31
I have a linux rig that runs pyrit  Cloud computing be fun. I don't have any real reason to upload in h265. If I did, I'd invest in a 10 core. The chip alone for x99 is over 800 right now. I don't think could justify that kind of money.

I found these Rent-a-Render pricing calculators.
Fox Render Farm
Rebus Render Farm

Being you know specifics of your Rig, you might be able to run the numbers, and give us a starting idea for owning vs. renting.


While searching, I came across  Render Rocket  that offered a "Weekend Warrior" package.  Gave me good chuckle.    But no "Spielberg" package...