【面白動画　Meshroom: 3D models from photos using free photogrammetry software】メッシュルーム 写真から3Dモデルを作成する、無料の写真測量ソフトウェア

このビデオでは、写真測量アプリケーション「Meshroom」について説明します。Meshroomは、Alice Visionフレームワークをベースに構築された、フリーでオープンソースのアプリケーションです。このアプリケーションは、ユーザーが一連の写真を3Dオブジェクトに変換することを可能にします。このソフトウェアは、3Dデジタル化技術の民主化を目的とした非営利団体「Alice Vision Association」によってサポートされています。Meshroomは、CUDAをサポートするNvidia GPUと、32GBのRAMを搭載したi7 PCまたはそれと同等の推奨スペックを必要とします。また、Nvidia GPUを搭載していない場合でも動作しますが、結果は低品質になります。このアプリケーションは、高レベルと低レベルの両方のユーザー・インターフェースを備えており、初心者から上級者まで利用することができます。Meshroomを使って、木製の象の3Dモデルを作成する様子をご紹介しています。

写真から3Dモデルを作成する「写真測量」について解説しています。Meshroomというソフトを使い、クラウドサーバーではなくローカルで処理を行うそうです。フォトグラメトリー処理の品質は、使用する画像の品質に依存し、最低でも20枚、詳細な対象物には数百枚が必要とされる。画像は、モーションブラーがなく、ノイズが少なく、被写界深度が深く、影や反射がなく、均一に照明されている必要があります。写真測量を成功させるためには、処理中にカメラの向きや焦点距離が変わらないようにする必要があります。Meshroomでフォトグラメトリー処理が完了すると、データは3Dビューアーで見ることができ、キャッシュフォルダにエクスポートされ、Windowsの3Dビューアーで開くことができるようになります。出力されるのは形状だけでなく、テクスチャデータも含まれます。

この対談では、Mesh Roomを使ったフォトグラメトリで3Dテクスチャオブジェクトを作成する方法と、Autodeskの無償プログラムMesh Mixerを使って出力をクリーンアップする方法を紹介します。まず、写真を3Dメッシュに変換するMesh Roomを使って3Dテクスチャ・オブジェクトを作成する方法を紹介し、次にMesh Mixerを使って出力をクリーン・アップする方法を実演します。次に、同じプロセスを石獅子に適用し、118枚の写真をMesh Roomにインポートして、Mesh Mixerで出力をクリーンアップします。講演者は最終的な結果に満足し、モデルのディテールレベルが非常に高いことに注目しています。

この対談では、写真測量ソフト（Meshroom）と静止画を使用して、ライオン像の3Dモデルを作成するプロセスを説明します。まず、Meshroomの出力をMeshmixerに取り込み、接地面との位置合わせ、余分なポリゴンの削除、拡大縮小、穴埋めなどの処理を行い、最終的に水密性の高いモデルを完成させる。次に、このモデルをSTLおよびOBJ形式でエクスポートし、3Dモデリングパッケージ（Lightwave）にインポートして、ゴールドシェーダーを適用し、アニメーションをレンダリングします。最後に、スライスソフトで3Dプリントし、充填、サンディング、塗装で仕上げます。最後に、Meshroomは写真測量に最適なソフトウェアであり、最終的なアウトプットに満足していると結論づけています。

welcome to another video from explaining
computers
this time we’re going to check out a
photogrammetry application called
meshroom
photogrammetry turns a series of
photographs into a 3d object
so for example a few years ago i used a
free photogrametry application called
autodesk 123d catch to turn this wooden
elephant into this 3d print
however since that time one 123d catch
has been replaced with an application
called recap pro that costs 340 dollars
a year
which gets us to meshroom which is a
free open source photogrammetry
application
so
let’s go and take
a closer look
right here we are on the website for
alice vision where we can download
mushroom
alice vision is a framework that
provides computer vision tools for
turning photographs into 3d models with
meshroom being a photogrammetry
application built around alice vision
all of this is supported by the alice
version association which is a
non-profit with the ambition to
democratize 3d digitization technologies
from photographs and here as you can see
users of meshroom can make a donation to
support the work of alice vision if they
wish
back on the main website if we click on
meshroom to download the software i’m
sure you’ve guessed that we can scroll
down and we will find two download links
one for windows and one for linux
also as it notes mesh room requires an
nvidia gpu with cuda support and here on
my test rig we’re running with a two
gigabyte gt1030
note that the broader recommended
hardware specification is an i7 pc or
equivalent with 32 gigabytes of ram
however here we’re running on a quad
core three gigahertz amd a8 3870 k with
16 gigabytes of ram on which mesh room
runs with no issues
finally on the hardware side it’s worth
noting that regardless of what’s stated
here it is possible to run mushroom
without an nvidia gpu and to do this we
go across some of the support pages and
we discover we can do it if we use draft
meshing and if we click on this you’ll
find that using this is very well
documented but everything to do with
mushroom is very well documented but if
you do use draft meshing you’ll get
lower quality results and it’s not
something i’ve tried out but i just
thought i’d mention this in case you
haven’t got an nvidia gpu
anyway let’s download the software so
i’ll click here on the windows link as
i’m currently on a windows 10 machine so
i’ll do that
and save the file which is 327 megabytes
in size and with the file downloaded if
we go across to the folder in which it’s
contained we need to extract the file so
we’ll extract there
and with the extraction completed no
more installation is required because
all we have to do is to open up the
folder where mushroom has ended up there
we are let’s just give ourselves a bit
more space on the screen and all we have
to do to run the package is to click on
the meshroom xc file and here we are in
this amazing software application
meshroom is incredibly well designed to
incorporate both a high level and a
low-level user interface
specifically at the top of the screen we
have the high-level ui which should
allow almost anybody to produce a 3d
object from a set of still images
and then at the bottom of the screen we
find the low-level ui which is node
based and should allow researchers and
advanced users to take complete control
of the photogrammetry process and i find
all the stuff going on down here very
interesting indeed but in this video
we’re going to stick with the high level
interface here at the top of the screen
and of course you want some
demonstrations i’m going to do two in
this video and i thought i’d start out
by creating a 3d model of the wooden
elephant like the one i previously
generated in autodesk 123d catch
for this i still have the set of 36
images that i took as i rotated around
the object to photograph it from all
possible angles so to start the process
here in mesh room the first thing we
need to do is to name our files i’m
going to do a file and i’m going to do a
save and we’ll go to a folder i created
called photogrammetry where you can see
i’ve already got a folder which contains
the elephant images and i’m going to
save this project with the name elephant
test
and it’s worth noting that it’s very
important to keep track of files and
folders and their locations when you’re
working in meshroom as output from the
program is saved in an auto-generated
folder called meshroom cache which will
sit alongside your project files
so with our file set up we’ll now load
in the images of the elephant so we’ll
go across to file and import images and
again back to photogrammetry here are
the images and we’ll just select all of
these and bring them in to the program
like this

and here they all are sitting there
elephant photographed from lots of
different angles and all we now have to
do to begin the photogrammetry process
is to click on start and unlike many
other photogrammetry applications
meshroom does all of its processing
locally it doesn’t rely on a cloud
server so this is going to take a rather
long time and we can see the progress
being made in two places we can see a
progress bar here at the top of the
screen this will take a while to move
across and we can also see down here in
this node view we can see which node is
doing something and again these will be
curled in as things move across
and whilst mushroom is getting on with
its task it’s very important to stress
that how well the photogrammetry process
will work depends on the quality of the
images it’s provided with
so just what do you need to do to
achieve photogrammetry success
well the first thing is to take a lot of
images an absolute minimum of 20 and to
get a really good scan of a detailed
object several hundred
photographs need to be sharp with no
motion blur as little noise as possible
and good depth of field so that no part
of the target object is out of focus
lighting is also critical and needs to
be as flat and diffuse as possible with
minimal shadows and no reflections
in an ideal world an object will be
photographed with lots of identical
cameras all taking a picture at exactly
the same point in time as can happen in
a booth like this one i saw at the eye
maker store in london
however most of us have to work with one
camera which is rotated around the
object and it’s worth stressing that
rotating the object itself can be
problematic if this changes how
different parts of it are lit
oh and two final tips firstly make sure
you don’t change the orientation of your
camera or in other words shoot
everything landscape or everything
portrait and also don’t change the focal
length of your lens or in other words
shoot every picture on the same zoom
setting
right
back in mesh room about 90 minutes has
passed and the photogrammetry process
has completed as we can see
and in the 3d viewer we’ve now got some
data and if i just zoom in on that a bit
in fact i’ll give us a bit more space on
the screen for it we’ll just reorganize
things a little bit like that there we
are and you can see what we have here
and i can move around with the mouse we
can see a representation of the camera
position for every one of the images if
i click on there you’ll see it actually
moves them around let’s give us a bit
more space to see more those as well
it’ll flick between cameras and we can
also zoom in on the elephant which you
can see is sitting in the middle of this
let’s just go in a bit closer on the
elephant there it is it’s upside down
that doesn’t doesn’t really matter we
can still cope with that we can change
the size of the points on the elephant
the ones that have been created there as
you might see coming up down there and
we can also change the size of the
cameras if it makes it easier to see
things
and something else we can do here which
is very interesting and we can click on
sync with image selection if i do that
we see the elephant based upon the image
that we’ve actually got selected over
here so it’s showing the point data
that’s coming from that particular image
which i i find fascinating so that’s a
that’s just a cool little feature here
anyway what we’re most interested in of
course is actually using this model in
other applications and you might be
thinking how do we export this model
from mesh room and there aren’t any
export functionality things here in the
file menu and the reason for that is
because mesh room saves all of its
output in its cache folder as i said
earlier so if we go across to the
photogrammetry folder you’ll see it’s
created a mushroom cache file here and
if we go into that and we go into
meshing you will see we have a very long
name folder there and inside that we
have a mesh object which here is for the
elephant and if we open this up
it’ll open up in the windows 3d viewer
and i’ll f9 to use the whole screen and
here we are
we have the elephant still upside down
and there’s a bit of a extra date a few
extra polygons and things around the top
but we can sort this out we can actually
remove these extra polygons and things
fairly easily as i’ll show you in the
next segment of the video
but just before we do that it’s worth
pointing out that here in this video all
i’m going to be looking at is the
geometry created by meshroom the actual
object we’re looking at on the screen
here without any color on it but
mushroom has also saved loads and loads
of texture data so if we take a look at
that let’s close down that and that and
go into text string in the cache and
again we found a file for this
particular project hello look there’s
the texture we open it up
and this is a png file containing all of
the textures all of the color data for the model

but i’m sure some of you are wondering
chris how did you get to this final
model from the initial meshroom output
so let’s go back to that output which
looked like this when i first loaded it
into meshmixer and the first thing i did
was to get it aligned with the ground
plane using edit and transform
and then i had to start getting rid of
all the extra polygons we don’t actually
need in the model
and to do that i use select and then
selected lasso and drew on the screen to
pick up polygons that weren’t needed
like that
and clicked on delete
and after a lot of rotating around and
selecting and deleting i got to this
point where there’s pretty much only the
lion left in the scene we’ve got a bit
of wall over here a bit of ground but i
soon got rid of those and ended up with
a line like this
next i dealt with scaling by going to
edit and transform again and i set the
height of a line to 105 millimeters
next i went to analysis and inspector to
try and make sure our object has got a
water tight mesh because if you want to
3d print something the geometry has got
to be watertight and as you can see
there’s a lot of issues with the
particular object but fortunately in
meshmixer there is an auto repair all
option so i just clicked on that
which fortunately sorted everything out
this said the liner still got a bit of a
problem because on the top of its head
it’s got this strange body of polygons a
strain sort of hat on the top of it and
this is because the actual stone line
that i photographed was over seven feet
tall so even lifting my camera above my
head i couldn’t take good pictures of
the top of it so once again what i
needed to do was to go to select and to
select all of these polygons and to
delete them
and with that process completed i ended
up here with a lion without all the
strange polygons on the top of its head
but was a hole in the top of its head
which needed to be filled in
and a hole like this could be filled in
in various ways but i went for the
simplest possible solution which was to
go to edit and make solid
and this as you can see does fill in the
hole if in a rather rudimentary fashion
although here we’ve lost a lot of
resolution a lot of detail in the object
with the current settings but we can fix
that by changing the settings for make
solid and i think i picked something
like that and then press the button at
the bottom of the screen which is
difficult to see here on my current
scaling settings but i can click it down
here
and by jingo i now ended up with a far
happier lion to which all i needed to do
was to go and edit and a plane cut
which gave a nice smooth base as we can
see here in the final object
and so the final thing to do was a
little bit of exporting first i exported
an stl file for 3d printing
and then secondly an obj file so i could
import the lion into the lightwave 3d
modeling package in which i make all of
my 3d animation and here i first
rendered the lion with a flat shade just
to see what it looks like and then i
gave it a gold shader and rendered out a
bit of animation spinning around the
model
and i think it’s come out very well
indeed
certainly a lot more time could have
been spent cleaning up the top of the
model and that internal cavity around
the back of the leg but there’s no doubt
at all we’ve proved here that meshroom
is a great piece of software for
performing photogrammetry for making a
3d object from a series of still
photographs
guess what i’m now 3d printing the
guardian lion
to achieve this i loaded the stl file
that we generated in mesh mixer into the
curious slicing software which in turn
produced a g-code file for the 3d
printer
obviously extruding the line into
physical reality will take many hours so
we’ll now let the printer get on with
its task and here we are with all the
support structures removed we have a
final 3d print
and it’s really great to hold in my hand
an object that i last saw in the
physical world as a two and a half ton
piece of stone
however the lion’s journey is not quite
complete yet as i’m going to do a little
filling and sanding and painting so that
we end up with a final model that can
guard something of importance on my desk
and i for one think this is a great
final output from our photogrammetry
process
as we’ve seen in this video meshroom is
an amazing piece of software that allows
us to turn something like this
into something like this
but now that’s it for another video if
you’ve enjoyed what you’ve seen here
please press that like button
if you haven’t subscribed please
subscribe and i hope to talk to you
again
very soon