Show Posts

This section allows you to view all posts made by this member. Note that you can only see posts made in areas you currently have access to.

Messages - CheeseAndJamSandwich

Pages: 1 ... 4 5 [6] 7 8 ... 11
The navigation rotation ball gizmo needs a reset button so that its red green blue lines are aligned when the model is in ortho view. I try to rotate the model in ortho and you think all is going well then it tumbles, if one can select the edge line of that little sphere it would rotate and keep its planar aspect the same, as such I then have to revisit the other two orthos and fix the alignment all over again., as there is no undo for any tumbling or rotation.
Again, rotating/resizing/moving the region rotates/moves the gizmo...
Align your grass with the grid, then align the region with the grid too, using 1 and 3... then align the region with the x and y with top view, 7.  You can then resize the region to place the gizmo in the middle of the object. Now the gizmo is will allow you to rotate is about the 3 axis.

I scale my model to a grid, as i know a distance between two rocks is 10 metres...  So i just set the grid to 10m...  then do the scaling first, placing the two rocks touching the grid, rotate to align them both with the grid, then scale, move, scale, move, scale, move, until you have the grid lines crossing on each rock...  then i reorientate the model to where it actually needs to be, and the same with the region.
Sadly the grid size is only in whole metres, or pixels... so a bit of a bummer if you're doing smaller stuff...

We really need the basic referencing sytem feature in the Standard version, as i've requested.
As this is a PITA.


File export model as obj name it, then up pops a panel with all sorts of settings, one is Precision and it shows 6 but no indication of what range there is and what the numbers mean,
Could we please get Tooltips added to MS's dialog boxes and preferences tabs?
They're very commonplace and very, very effective at giving the user a description of the feature, variable, setting, etc, and suggesting values, when to and not to enable something...
This would help so, so much. 
And it would ultimately reduce the number of forum posts!  :P 8)

The issue is, that unless you have the Professional version, we can't enter references, control points, markers, so MS can't know which way is up for whatever object you scan...  Tho i think if your imagagery is from a drone, i think it does takes its orientations info from the pics?  Correct???

I've made a Feature Request asking for a very basic referencing system that'll allow us to place at least 3 markers, control points down, which then allows us to correctly set the orientation and scale of the model.  There's so much to MS Pro's referencing system it seems, that the subset i'm requesting shouldn't tread on the toes of the Pro version, and it's use in the industries that demand very high level of georeferencing, etc...
Having 3 or 4 basic xyz control points would just make everything really easy...

All i do after alignment is to rotate the model relative to the grid, in ortho view (5), pressing 1, 3, 7 to get the views from the top front and sides (ignore the left/right, they're actually the other way around, but it doesn't matter really).
I then scale it to the grid (set grid to 5m in Prefs), using features i know the distance of.
Then rotate and resize the region to match the model.  The ball, etc then aligns with the z-up of the model.
For me, the trackball is just waaay to small, so it makes rotating the model around really twitchy and cumbersome, compared to what i was use to with good ol' CATIA V4, that i once used a lot.

So yeah, Metashape could do with some work on the 3D navigation...  And if we had a choice of styles in the Preferences, with a parameter or two when required, it'd just keep everyone happy, and each user would get that bit more out of MS...  Helping MS sell more copies.

Feature Requests / Re: Align/snap model and region to axes
« on: October 17, 2021, 05:39:43 PM »
You're not alone.
I posted this a few days ago...

If we could get a very basic reference system in the Standard version, allowing us to put down a minimum of 3 markers, control points, that we know the positions of, then the models would snap to the exact correct orientation and scale.
Leave out all the georeferencing stuff the the big boys use for scanning mine sites, etc. that needs Metashape Professional for...

So for your turntable work, you could put down 3 point around the rim of the turntable, that you know the x,y coordinates for, in milometers perhaps, and then zet z to zero.  This make the model's orientation and scale perfect.

Every time i create a model, which is derived from gopro images, underwater, the model is never, ever in the correct orientation or the correct scale... and i have to do the same workaround procedure to fix it.

It would be awesome to get some basic tools to make this work.

*irony off*

(Such software with so many complex settings really deserves better documentation.)
This is a very, very important point.

We're lucky that we do have a these forums, and Alexey, etc, helping out in most posts... But the documentation is OK in most areas, and 'lacking' in many.
This is a complex application doing so many different things, with lots and lots of features, that help us in so many ways...  But so little is written about some of them.

We're also blessed with an astoundingly easy workflow that works so, so well out of the box, so we can get epic results with almost no effort or experience with the software...  But the devil is in the details...  When things don't quite work as hoped, when we need to go the extra miles to get the very accurate results demanded by the job at hand, then we have to get dirty and get technical, using all the little special  features that do very specific things...  Part of it is knowing how to use them... bit a big part of it is knowing that they exist in the first place!!!

As with any complex software like this, if you have good documentation, and a good user community, then the software can be very easy to use and get the job done, meaning that it'll get adopted by more and more of the industries that need it.
The easiest to use, most well documented, most supported software will win over the complex unsupported ones, even if they're match in ability.

Some things that would make Metashape better:
  • Tooltips:  Tooltips are amazing!  And familiar to us all.  MS uses them already for the toolbar buttons, but it could use them a LOT MORE?  Tooltips be added to every setting in the preferences dialogs (which is a common place for them) defining the setting and suggesting values.
    Could they also be added for all the menu items?  Not just the toolbar buttons?  Not that common, but still very, very desirable.
    The tool tips should be an explanation of the feature, and if possible, with perhaps a few notes on applications, suggested values, etc.
    As tooltips appear after a delay, even a long explanation, or lots of details about it, would not affect the usage, experience by the user, it won't get in the way.  If you pause, you get a good explanation.  If you know what you're doing, they proably don't get enough time to appear.
    Flesh these tooltips out, and enable them everywhere!
  • Example scenarios:  In the manual, it would be great if a lot of the features had examples of where it could be used, and how it would be used for different uses of MS... So there would be examples for aerial/drone work, examples for turntable rigs, for interiors, etc, etc. where the settings might be different for the different uses.
  • Crowd-sourced explanations and examples:  Perhaps the users that have gained a lot of experience with MS could submit their settings and insight about various features, say as concise case studies.
    It could even be that each time Alexey fleshes out an explanation of a feature for us, after a forum user posts, then we could take that explanation and submit it in a 'Tooltip' format and a revised Manual entry.  The documentation would then grow as a when features are asked about and explained.
  • Workflows:  One of the most useful, educational, and even reassuring documents, is the USGS Processing Workflow.  Probably one of the most valuable resources to a lot of us.  It tells us what, when and why, to get the outputs they needed.
    If we could get a collection of these, submitted by us users, for all of our different use cases, then that would be absolutely amazing, so useful.
    Each workflow could then be critiqued by us peers and the devs, and then perhaps updated as we learn new things, new features come along.  The author optimises their workflow, we all learn from their workflow.
    Having a few different workflows for your type of use case would be great.  Some for drone surveys, some for underwater surveys, some for scanning archaeological sites, or museum specimens on turntables, scanning cars, etc, etc, etc...
    This would allow any new user to get up to a reasonable speed, very, very quickly!

Fleshing out good quality documentation, manuals, etc is not easy, it's very time consuming, and needs maintaining... The devs are busy enough coding new features and making MS faster... But it's very, very important...  If it can be improved, it absolutely will help us all...  And the easier to use software will ultimately sell more copies.

General / Re: What needs doing to be able to use Reduce Overlap command ?
« on: October 17, 2021, 04:02:05 PM »
If the actual scene is not too complex, you can generate a rough model from the sparse point cloud and run Reduce Overlap command.

How do you define 'not too complex'???

I create the sparse point cloud derived mesh with the default setting, and it creates it very, very quickly!!!  Even with the sparse point cloud generated from 20,000 cameras...
Should i, could i up the settings so the sparse point cloud mesh is even higher quality?  Otherwise i have to wait 24 hours for it to create a mesh from the all 20,000 cameras...
What kind of complex scene shouldn't use the sparse point cloud?

General / Re: What needs doing to be able to use Reduce Overlap command ?
« on: October 17, 2021, 03:55:47 PM »
I think the Reduce Overlap's Surface Coverage value is the number of photos where a point/feature is seen in...  Correct Alexey?
I use 9...  so a point on a rock, that MS has recognised and put a tie point on, is seen in up to 9 different photos, from 9 different angles...
The nature of my scanning, swimming around with GoPros taking 2s timelapse photos, means that i always land up with too much overlap for each photo, often overlapping tracks, so too many images.  Using the Reduce Overlap's Surface Coverage value set to 9 usually disables about 1/3 of my cameras.  And removing these absolutely speeds up the depth maps and meshing heaps!!!

My procedure:
Align all photos.
Try to align the unaligned (reset & align selected)
Gradual selection Recon & Proj set to only 15, it removes a lot of the weak points.
optimise cameras.
Create mesh from sparse cloud (VERY QUICK!)
Reduce Alignment set to 9
Photo Pane, sort by Enabled, select all the unticked, remove selected photos.  This deletes 1/3 of the cameras.
Optimise cameras again
Create mesh from Depth maps...

Just save your model before you run the Reduce overlap command, and then just reload it trying different values, until you find one that's removing just the right amount.

The ball is dictated by the region.
It took me ages to work this out myself, lol!
And they also dictate where the centre of rotation and height of the model is if you tell it to do the default horizontal orbit animation too!!!  It look me ages to work this out too!

Rotate the model to the correct orientation, move it's z so that it's roughly sitting on the grid, z=0, then rotate and move the region to match the model... the ball will be in the very centre of the region.  The ball's axis, rings will then mover it correctly.

The other thing that's weird as, is that if there's things off to the sides, away from the model itself, like animation cameras and tracks, it completely screws up Default View, pressing 0...  It's related to the extents of all the data and objects within the project i think...  Not the extents of the points, mesh, etc...  Even then, it doesn't zoom out to show the extents... it just zooms to an illogical view...  Confusing af.

You are correct it seems.....
I just checked with one of my models, a sunken pontoon, in top view, 7, i rotate it until the front of the pontoon is at the bottom of the screen...  then when i press ctrl 1, the front, the bow, is presented correctly, and if i press 3, it shows me the left hand side, port side, of the pontoon!  ctrl 3 shows me the right hand side, the starboard side!  1 correctly shows the back, the stern... 
This is indeed the wrong way to what should be logical...   :o
Right = Left, and Left = Right...  Doh!
I guess we just ignore what it says it is, and just know that 3 is left, and ctrl 3 is right!!!  ;D

Feature Requests / Re: Close Holes - Within Selection
« on: October 16, 2021, 10:54:57 AM »

Feature Requests / Re: Scale function for standard version
« on: October 16, 2021, 10:52:27 AM »
There is really one basic feature in the standard version.
This is really important feature to prepare for 3D printing.

Please add simple scale function to the Photoscan Standard version.
I've successfully used Alexey's workaround method, but it's fiddly and time consuming to get right.

I've made a feature request for Metashape Standard to get a basic reference system, that allows us to put down at least 3 markers, with know x,y,z coordinates, such that the model would then be perfectly positioned in 3D space, and, as you need, correctly scaled.

My feature request:

Also a larger rotation widget for finer control, and one which you can recenter by double clicking as you do with recentering the point of view rotation in navigation mode would complete everything i want for Christmas... this year at least.
I've also requested a larger trackball.  As i find it's tiny size makes it far to twitchy when rotating.  If we could edit its ratio to the viewport size ourselves, then we'd all get the perfect Christmas present!
I'm an old Catia V4 user, and it's trackball was perfect.  MS' one i struggle with still.....

Feature Requests / Re: Additional optional grids for yz, xz?
« on: October 16, 2021, 10:37:03 AM »

In preferences, we can set which plane the grid is on: xy, xz, yz, and also it's cell size and unit, etc.

But for sure, it'd be nice to be able to toggle the other two planes on and off too.
Or perhaps a key to switch planes, such that it would cycle between xy, xz and yz.  Perhaps momentarily flashing up the plane the grid is now on, or including it in the Info, when the grid is on?

In Metashape Standard, we don't have any way defining the orientating or scaling the model, other than manually... We don't have the ability to add markers with known coordinates.

We can only orientate and scale it by using the workaround method of setting the grid to, say, 5m, 10CM, 10', 12", whatever, and then manually scaling the model or features on it, roughly to the grid, and rotating into into the correct oriendation horizontally, vertically and rotating it so that north is up perhaps.

Could Metashape Standard have a very basic reference system? A tiny subset of what Professional has?

If we could at least have the ability to place down 3 markers, with x,y,z coordinates, this would kinematically define the model's position perfectly in space, and scale it.

Perhaps we could only use Euclidean/Cartesian coordinates, x, y, z
So no access to georeferenced coordinates, WGS84, etc, etc...
3 points is the minimum to define something's position & scale in 3D space, but perhaps we could get 4, or 5 points, such that it would help with camera calibration issues, bowing, dishing of the model? But a lot more code i guess.

Just being able to also add 3 markers to our photos before alignment would be amazing, saving a lot of work.

Feature Requests / Re: Edit size of 'Show Cameras' rectangles
« on: October 15, 2021, 08:56:49 PM »
Hello CheeseAndJamSandwich,
As for the color, it cannot be changed at the moment, but as a workaround, you can select all cameras - and the placeholders will be shown in red color.
Yes, that works for now, thanks!

Pages: 1 ... 4 5 [6] 7 8 ... 11