POP 3 single shot , is this normal?

Tried out the single-shot method today and was quite surprised to see how rough a frame is. The high-to-low distance is as much as 0.2 mm ! The object is a piece of plywood with smooth surface painted in matt white hence perfect for scanning. The exposure was controlled so that the red patches in the depth camera view of Revo Scan just disappeared.

Is this normal ?

Single-shot one frame

Single-shot 10 frames. Neither the scanner nor the object has moved.

Roughness measured in Fusion 360

Continuous scanning

Look a single shot frame whole scan with multiple single frames .

I have seen a post saying that it will be sufficient to take 6 frames to build the model of a simple cube and that is the best model as there is zero over scanning hence smoothest surface. Not sure if I have understood it incorrectly but the result I got from my POP3 is not quite supportive of that.

I will check it for you with POP3 and see the results if there are any issues with the firmware/software so you know if that is your device or not .

You mentioned early the issue with calibration so that may be the case , but to be sure I need to replicate that bug myself .

Thanks !

1 Like

I have the exact same problem in single-shot mode. :neutral_face:
What could this be related to?

fusing method - advanced

fusing method - standart

@PUTV Could you please show what a one shot looks like in single shot mode on your pop3?

Hi @mikinik usually it is related to an angle , some objects needs different angle while scanning , especially those that have hard edge .
Try to scan it at a different angle of the scanner to see improvement.

The lines are results of the pattern projected at the specific angle , flatter surfaces change accuracy since the pattern get from sharp to blurred with the distance .

I get this with every scanner - single shot mode scans are waaaaay rougher than continuous mode.
Not exactly sure where I heard it but I too thought that single shot scanning is the way to go for max details - but all the tests I’ve done on Mini, Pop2, Range and Miraco show this not to be true.
The scans Fuse quickly, maybe the photo texture quality is better, but the surface itself is always very grainy. The more shots you take, the better it looks - but at some point you come to a conclusion what’s the point - might as well scan continuously and get a better scan in shorter time.
Would be nice if Revopoint did video comparing both scanning methods - showing the resulting meshes from both - maybe there is something simple we’re all missing.


Also one thing I found strange. Even though I’m doing a single shot scan, it seems Revoscan expects the shots to be in logical sequence. It’s not enough to have a meshe overlap of 40-50%. The overlap has to be in the direction the previous shots were moving.
If I just stop what I’ve been doing and just walk around the moddel from the other side and take a shot - I could have even 90% of mesh overlap on a shot like this, Revoscan will still fail to match the data.

1 Like

I would disagree, both was done using single shot mode between 32-90 frames

It’s not about the overlap it is about the features , no features no alignment .

1 Like

I stand by what I said.
The shots have to be in a logical sequence - as if frames from a continuous sequence scan.
You can’t just jump around the model, no matter how much overlap you have and features.
Imagine you start the scan face front and then you slowly go to the right side, shot by shot.
When you reach the ear you want to do the left side.
You can’t just go back to face front and this time go to the left.
You have to go shot by shot from the right most ear slowly back to the face front and only then continue to the left.
It’s as if the shot matching is only with the previous shot, not the entire point cloud scanned until now.

90 frames is a hell of lot for single shot.
That’s like 9 seconds of continuous scanning, and I’m pretty sure you can scan these models in 9 seconds (assuming they are not scanned from the other side).
I guess we kind of expected the single shot mode to work more like photogrammetry.

It just occurred to me.
Miraco’s camera sensor is not very good - even in full light the photos are grainy.
It’s logical to assume that the IR camera sensor is of similar quality.
That means every single depth frame is also grainy.
That explains why the single scans are so noisy.
With continuous mode we get many frames of the exact same area - a sort of overscan of the area - an so it’s easy for the software to average out the grain from the depth camera.
In single shot mode, when there is less or almost no overscan - there is no easy way to remove the noise but leave the detail.
As you suggested, we can create a lot more single frames, creating our own overscan data - but then again that’s what the continuous mode is for.
I’m finding no real use for the single shot scan mode:

  • I either take as much shots as I would with photogrammetry - then I get noisy scans
  • or I take way more shots - but then using continuous mode is just faster and easier
1 Like

90 frames is a hell of a lot of single shots?
Compared to 700 for 2 angles only ?
You need to scan all angles not just front one angle , so 30 frames per angle in place of 350 per one angle .

And no it was full scan and I jumped from one side to another if areas was missed , this particular object was impossible to scan with continues mode do to the wings as the distance between the body and the tip of the wings was too great , so I had to jump on the wings separately after scanning the center body .

If course you sharing your own experience here … this don’t means if you can’t do it, everyone else can’t either.


As I stated above I would be not able to scan the object above without single frame due to the wings , there is lots of usages for a single shot usage , it is not a photogrammetry so the photogrammetry workflow should be not used .
You still need specific features to be available in each shot to be able to continue scanning , proper consistent distance as well to keep the proper accuracy on the whole length .

I would definitely not use single shot mode for scanning objects that move like for example humans Faces/Body , objects without unique features like bloc, cubes, spheres and all primitive objects that don’t have specific features .

Single shot or continuous mode still requires the same rules .

Damn you.
I have to redo my tests… again.
Honestly, it’s enough that it works for one person.
If it does, it means everyone else is just doing it wrong :slight_smile:
90 shots for both sides of the models is alright I guess.
But I’ve been playing with my daughter this weekend scanning ourselves and we get a very good face scan with an average of 200-300 frames in continuous mode.
Hence my reaction - needing half as much frames in single shot mode is not really that appealing.
But I do agree - scanning the bugs from both sides with wings would be impossible in continuous mode without some auxilary objects around it. Single shot mode seems like good idea here allowing us to take our time selecting the exact angles to shoot :thinking:

1 Like

I did loser tracking each time the wings hit the sides in continue mode and the accuracy was not as sharp as the main body , single shot fixed it since I moved from left to right in place of rotate it on turntable so I could capture everything at once with only 90 frames .