Does Mavic Shoot Video or Watercolor Paintings? Don’t settle. Get Involved!

[borislip]

Can you clarify? When you say ".... you get this". What do you get?

This - watercolor effect. The watercolor effect is nothing but noise reduction along with compression applied to a relatively even areas without much details. Which is exactly what trees turn into when they are out of focus.

The continuous autofocus they have added is terrible to say the least. It looks like it attempts to see if the center is out of focus, with many false positives, and running the entire focusing process from the beginning, which is slow, and causes pretty much everything to go out of focus until it finally regains focus again. And this happens quite often. This is nowhere near a true continuous focus, which should track the focus and do the necessary adjustments.

Speaking about decreasing resolution also decreasing the CPU load or something - this sounds rather unlikely. You'd see frame drops if that wss the case, either with CPU or HW codec. But what decreasing resolution and frame rate CAN do is make the compression less aggressive, assuming the output bitrate is the same (which, strangely, seems to be the case - 60mpbs, on any settings).

In any case, there isn't really that much improvement between "watercolor" and just plain "out of focus" trees and similar kind of objects.

The bottom line is, disable the new, destructive, AFC function. And FOCUS, FOCUS, FOCUS. Preferably somewhere close enough to the hyperfocal distance. But don't keep re-focusing needlessly either, or you will do the same damage to the image the AFC doed.

 

[RitterRunkel]

If only ... focus would be the matter. While it's important to focus correctly, I'm afraid, it isn't the issue here.

 

[jjmucker]

If only ... focus would be the matter. While it's important to focus correctly, I'm afraid, it isn't the issue here.

Agreed. This thread is getting a little lost here.

 

[cardplayer71]

Agreed. This thread is getting a little lost here.

I disagree. I saw very similar artifacts in my videos due to focus hunting. They stopped when I went manual.

 

[borislip]

I disagree. I saw very similar artifacts in my videos due to focus hunting. They stopped when I went manual.

Exactly. Maybe some Mavic cameras are just plain defective, and generally aren't sharp enough while fully focused. This should probably be tested by taking a well focused raw photo on such a Mavic.

 

[RitterRunkel]

I'm fully satisfied with sharpness and detail in the RAW photos from my Mavic. JPEG is crap - of course, since I already know this from photography. No model of all brands/manufacturers I own and have seen is able to create fine JPEGs out of cam. It's acceptable as a quick result. And of course, 8 bit per channel, and most often 4:2:0/4:2:2, etc. it's JPEG and not RAW, so I don't expect to get fine detail here.

But the denoising (watercolour glitch, whatever) issue in the video image of the Mavic still annoys me. Though I'm quite sure it's not a focus issue (e.g. regions with thin sharp lines wouldn't occure), I'll switch to manual focus next time. But I don't expect this to be successful.

 

[sirrraj]

Speaking about decreasing resolution also decreasing the CPU load or something - this sounds rather unlikely. You'd see frame drops if that wss the case, either with CPU or HW codec. But what decreasing resolution and frame rate CAN do is make the compression less aggressive, assuming the output bitrate is the same (which, strangely, seems to be the case - 60mpbs, on any settings).

I was wondering if 4K AND 2.7K really are both 60Mbps. I just checked my videos and all 4K files are 60Mbps and all 2.7K files are only 45Mbps.

 

[RitterRunkel]

I didn't try 2.7K yet. I'm using 4K. At the moment at 24 fps - to leave more bits per image for encoding ... and therefore be able to observe/analyse the watercolor issue more independent from compression.

But I'm curious: Does 2.7K result in a cropped or a resized image?

 

[4wd]

2.7K comes in at 2720 x 1536.
It's a bit cumbersome if you want to aim for youtube's 1440 setting it might be best to render at that rather than full size and let them do it.
I've not been convinced recording at 2.7 makes any difference except it's easier and quicker editing later - because the bitrate drops to 45 from 60 so compression must be similar so far as I can see.

 

[RitterRunkel]

With 1536p the amount of pixels is 50%. So with 45 Mbps you'd still achieve a better compression ... But it'd be sad if one was not able to set it to 60Mbps as well. My guess: while you might avoid some artefacts, it's better to watch and save 4K. Especially because you won't loose more color resolution by recording 2.7K in 4:2:0 subsampling.

 

[Tippocta]

Sent mine in for review/repair on Feb 15 and finally got an answer today. They are requesting that I pay for a new gimble and camera which was not damaged by me in any way.

This is ridiculous!

Anyone else encounter this with positive results? I would appreciate any info that you may have.

Thanks.

Dear Customer, Unfortunately, damage that is not caused by a product malfunction or is out of the warranty period is not covered by DJI aftersales policy. Damage not caused by product malfunction. We'll either repair it or replace it with a product that's new or equivalent to new in both performance and reliability after payment has been received. For more information, please visit (After-Sales Service Policies - DJI) - DJI North America


Item Quantity Unit Price（USD） Total Price（USD）
Mavic Pro Gimbal & Camera 1 179.00 179.00
Middle Frame 1 4.00 4.00
Top Cover 1 1.00 1.00
Service Charge 2 65.00/H 130.00
Freight: 0.00
Total Amount: 314.00
DJI Care deduction： 0.00
Total Payment: 314.00

 

[JPMavic]

& Back to the thread .. Here are my latest findings. My Mavic was definitely affected by my attempts to get the shutter speed down to 2 x Frame rate using ND filters. Now shooting with any filter that will keep the SS beneath 1000th. Anything near 60th causes issues. Just ordered a UV filter to protect the lens and will probably fly a lot with that alone. Colour is back to Art. -1,-1,-1. Sharpness at -2 became unpredictable. I could sharpen in post but not everything in the shot would come up equally. In regard to focus. With my Mavic this isn't a focus issue. This is a software or firmware issue. The one thing I have started doing is getting focussed first then staying there. Turn off Auto Focus. (I'm experimenting with setting manual focus Vs Auto on the furthest object but that's for another thread). The important thing for me is minimising the amount of commands you ask the Mavic to deal to as few as possible during each flight, and refocussing is one of those commands. I'm convinced that during the flight the processor becomes overwhelmed and the quality of the image degrades. (I've not read any threads which say that the image was terrible at the beginning of the flight and then got better). Therefore on fee paying jobs what I'll be doing from now on is landing half way through each battery. Switching off, booting up, toggle NTSC/PAL/NTSC. If I'm right this should pull the flush on the cache. This is what I think the key is.

 

[dangerm0us]

& Back to the thread .. Here are my latest findings. My Mavic was definitely affected by my attempts to get the shutter speed down to 2 x Frame rate using ND filters. Now shooting with any filter that will keep the SS beneath 1000th. Anything near 60th causes issues. Just ordered a UV filter to protect the lens and will probably fly a lot with that alone. Colour is back to Art. -1,-1,-1. Sharpness at -2 became unpredictable. I could sharpen in post but not everything in the shot would come up equally. In regard to focus. With my Mavic this isn't a focus issue. This is a software or firmware issue. The one thing I have started doing is getting focussed first then staying there. Turn off Auto Focus. (I'm experimenting with setting manual focus Vs Auto on the furthest object but that's for another thread). The important thing for me is minimising the amount of commands you ask the Mavic to deal to as few as possible during each flight, and refocussing is one of those commands. I'm convinced that during the flight the processor becomes overwhelmed and the quality of the image degrades. (I've not read any threads which say that the image was terrible at the beginning of the flight and then got better). Therefore on fee paying jobs what I'll be doing from now on is landing half way through each battery. Switching off, booting up, toggle NTSC/PAL/NTSC. If I'm right this should pull the flush on the cache. This is what I think the key is.

Glad we're back OT.

I agree with your observations and you have done more work than I have here. But a couple of things: my Mavic has this problem at startup, doesn't seem to happen midflight. So for me my issue is the same as yours, the use of filters and a low shutter 'causes' the issue. But I had started thinking there was a low threshold of light entering the sensor that causes it, at least for me. I don't see the issue when I'm not using a filter.

Here's my ND16 test (I haven't done any other tests properly yet). Sorry this footage is unedited apart from upacaling to 60FPS so it's a boring and uncinematic flight, but you can see the issue with the filter. Although I have no comparison footage to share, I don't have the issue without the filter. The trees are mush.

 

[JPMavic]

Glad we're back OT.

I agree with your observations and you have done more work than I have here. But a couple of things: my Mavic has this problem at startup, doesn't seem to happen midflight. So for me my issue is the same as yours, the use of filters and a low shutter 'causes' the issue. But I had started thinking there was a low threshold of light entering the sensor that causes it, at least for me. I don't see the issue when I'm not using a filter.

Here's my ND16 test (I haven't done any other tests properly yet). Sorry this footage is unedited apart from upacaling to 60FPS so it's a boring and uncinematic flight, but you can see the issue with the filter. Although I have no comparison footage to share, I don't have the issue without the filter. The trees are mush.

My ND8 filter would cause issues on anything but the brightest day so the ND16 would be out of the question for my Mavic. The mush in the trees is the best indication that the processer is struggling. Can you recall what shutter speed you were getting with the ND16? For me it needs to be a lot higher than 60th. If it doesn't happen with no filter? Maybe that's the answer right there. It would be interesting to know if the problems are restricted to those using manual exposure? When the cameras on Auto it seems to be able to sort itself out

 

[dangerm0us]

Perhaps I was overzealous by using the ND16. I do think that footage is a little dark. I did use the Polar Pro app though. I had a shutter speed of 1250 with no filter, and with a target shutter of 1/60 the Polar Pro IOS App advised that ND16 was the correct filter to use. p.s. I havent tried an ND Filter with Auto on! Will try that and see what happens.

 

[RitterRunkel]

It's eye-catching how many of us use manual slow shutter speeds in combination with filters. I'll take off my ND-PL for a while.

My ND8 filter would cause issues on anything but the brightest day so the ND16 would be out of the question for my Mavic. The mush in the trees is the best indication that the processer is struggling. Can you recall what shutter speed you were getting with the ND16? For me it needs to be a lot higher than 60th. If it doesn't happen with no filter? Maybe that's the answer right there. It would be interesting to know if the problems are restricted to those using manual exposure? When the cameras on Auto it seems to be able to sort itself out

Is the bold sentence related to the filter usage? Why should the processor work differently with filters? Of course you get more mush with trees close to the copter while moving. But other than that?

Still all I can imagine is that DJI is applying a logic like "slow shutter speed means dark scene outside, so increase denoising", if slow shutter speeds turn out to be the reason for mush in trees.

 

[daithi23]

Why should the processor work differently with filters?

Presumably because more time is spent capturing each frame of video, that is 1/60th of a second is what most aim for but without an ND filter it could be a quarter of that time.

 

[AyeYo]

Why do people shoot negative sharpness and then wonder why they have non-sharp images? That always blows my mind. I'm still issue free at 0-0-0.

The AFC has indeed ruined video though and since my GO4 app refused to update, I seem to be stuck with it because 4.0.2 doesn't have the option to turn it off. It's amazing that a company the size of DJI can't be bothered to actually test these changes before rolling them out.

 

[JPMavic]

Why do people shoot negative sharpness and then wonder why they have non-sharp images? That always blows my mind. I'm still issue free at 0-0-0.

The AFC has indeed ruined video though and since my GO4 app refused to update, I seem to be stuck with it because 4.0.2 doesn't have the option to turn it off. It's amazing that a company the size of DJI can't be bothered to actually test these changes before rolling them out.

Wrong thread I'm afraid. No one on this one is reporting an issue with lack of sharpness.

 

[JPMavic]

It's eye-catching how many of us use manual slow shutter speeds in combination with filters. I'll take off my ND-PL for a while.


Is the bold sentence related to the filter usage? Why should the processor work differently with filters? Of course you get more mush with trees close to the copter while moving. But other than that?

Still all I can imagine is that DJI is applying a logic like "slow shutter speed means dark scene outside, so increase denoising", if slow shutter speeds turn out to be the reason for mush in trees.

Why should the processor work differently with filters? I have no idea but reducing the ND strength and forgetting about the 2 x framerate Etc. Is definitely working for the camera on my Mavic.