Site is active again, info concerning photogrammetry
by Ron Kapaun · 11/03/2014 (6:03 pm) · 48 comments
Howdy!
Long time, no blogs!
Well, many of you noted (and emailed) that my site was down. Simple fact, the site used WAY more bandwidth than what I paid for... Sorry about the downtime. However, the site is back up, and it's all the 'same stuff for free'. However, I will be making adjustments to the downloads. These adjustments will include removing the demo levels and 'shrinking' the overall size of the zip files. I will include .pdf files that will include common 'errors' and such and help new users install the content properly. This is a good opportunity to 'revamp' the packs to make them more user friendly and with luck, help new folks produce good quality visuals in their games and test levels.
WHAT THE H*LL HAVE YOU BEEN DOING!
Well, as many of you know, I have been dealing with T3D more as a hobby again. HOWEVER, I have been 'innovating' and coming up with a number of interesting NEW art techniques and processes based on.... well, I am not going to into it right now but, if any of you played or looked at The Vanishing of Ethan Carter, you will have at least an idea of what I have been working on.
So your site was 'Crashed' WTF???
Well, I am currently not making ANYTHING from these free models etc. So, this means I have to 'work' at a real job. I have been luck enough to actually pick up a programming job (full time) using Visual Basic, and a freaking HOST of MS products (Access, Excel, Word, Sharepoint, etc.) I am basically creating a number of automation systems using VBA and MS to basically do with one person, what a year ago took 2 or 3 people. (This is how bad the 'military' budget has gotten.) Anyway, I spend 45+ hours a week, sleeping in my car and working on these projects so, I do not get a whole heck of a lot of time to work my web stuff or my T3D dev. I hope this improves shortly and I am sorry for the MAJOR delays. Anyway, I will post more when I have some good stuff to show off (I have some of course, I just have not made a video or anything yet :-).
At last, I wanna say FAREWELL to a number of peeps that have 'left' the Garage Games Community. Will Zetter et all thanks for your work and time. I wish you all well. (there are more than a few and they will all be missed!)
With that, I will leave you all for now..... HOWEVER!
MORE TO FOLLOW
Ron
Long time, no blogs!
Well, many of you noted (and emailed) that my site was down. Simple fact, the site used WAY more bandwidth than what I paid for... Sorry about the downtime. However, the site is back up, and it's all the 'same stuff for free'. However, I will be making adjustments to the downloads. These adjustments will include removing the demo levels and 'shrinking' the overall size of the zip files. I will include .pdf files that will include common 'errors' and such and help new users install the content properly. This is a good opportunity to 'revamp' the packs to make them more user friendly and with luck, help new folks produce good quality visuals in their games and test levels.
WHAT THE H*LL HAVE YOU BEEN DOING!
Well, as many of you know, I have been dealing with T3D more as a hobby again. HOWEVER, I have been 'innovating' and coming up with a number of interesting NEW art techniques and processes based on.... well, I am not going to into it right now but, if any of you played or looked at The Vanishing of Ethan Carter, you will have at least an idea of what I have been working on.
So your site was 'Crashed' WTF???
Well, I am currently not making ANYTHING from these free models etc. So, this means I have to 'work' at a real job. I have been luck enough to actually pick up a programming job (full time) using Visual Basic, and a freaking HOST of MS products (Access, Excel, Word, Sharepoint, etc.) I am basically creating a number of automation systems using VBA and MS to basically do with one person, what a year ago took 2 or 3 people. (This is how bad the 'military' budget has gotten.) Anyway, I spend 45+ hours a week, sleeping in my car and working on these projects so, I do not get a whole heck of a lot of time to work my web stuff or my T3D dev. I hope this improves shortly and I am sorry for the MAJOR delays. Anyway, I will post more when I have some good stuff to show off (I have some of course, I just have not made a video or anything yet :-).
At last, I wanna say FAREWELL to a number of peeps that have 'left' the Garage Games Community. Will Zetter et all thanks for your work and time. I wish you all well. (there are more than a few and they will all be missed!)
With that, I will leave you all for now..... HOWEVER!
MORE TO FOLLOW
Ron
About the author
Guy that has been with Torque Game Engines since... well, since here was a Torque Game Engine. (V12 and beyond).
#22
That'd be cool. Alternatively, I spend a large portion of my day in the GG chat, so it'll be easy to find me in there too :)
11/10/2014 (7:11 am)
@RonThat'd be cool. Alternatively, I spend a large portion of my day in the GG chat, so it'll be easy to find me in there too :)
#23
Anyone knows if there are already open source alternatives for this technique? Otherwise what is the cheapest software/hardware for it?
11/10/2014 (9:47 am)
This Photoscaning sounds too good to be true. Someone in the irc chat was also playing around with this, he send his model to me and I was able to relatively quickly bake it to a low poly version. As I expected there is a hook on this method, since you have to remodel a low poly version, but from then the scan delivers you all the texturing and normals maybe also. So for small objects this may be efficient, but what is with larger objects?Anyone knows if there are already open source alternatives for this technique? Otherwise what is the cheapest software/hardware for it?
#24
@Duion, I was skeptical at first too but, I have managed to bring a VERY detailed model into my blender pipeline and export a T3D compatible model that has LODs and proper MIP maps, basically everything, in just a few dozen hours from 'I am a newbie to this photogametry thing'. ADDITIONALLY, you are correct, diffuse, normals, heck even ambient occulusion and be baked in. The most impressive thing was the integration into my 'newish' Blender pipeline. Where I can bake in 'edge mapping' and a ton of other things. I so need to put out a new video showing a comparison. I am thinking a 'reboot' of the conifer pack using using these new techniques versus the pack everyone knows.
In my experience so far, This IS NOT FAST, it's NOT easy and I am getting by because I have A LOT of experience with 3D. Even my 'big machine' 16 gigs ram, plus dual nvidia 780 ti GFX cards has been GRINDING away at some of these scans. Granted, I am a newbie at this and I see and understand things even better than I did last night. I am planning a video to explain it all but, it's not a 'beginner' task by any means.
More to follow.
Ron
11/10/2014 (10:42 am)
Well, I have purchased the standard edition. It took me exactly 2 attempts to see the value in this method. It does not have the added expense of detailed line lasers and "additional' cameras and I have to admit, so far I am very impressed. I am using my 'standard' GE X400 (14mpx) camera and I have to admit I was not headed down the wrong path. (Though i still see a great deal of value in SLS v. laser scanning.) Jeff, we will defiantly talk!@Duion, I was skeptical at first too but, I have managed to bring a VERY detailed model into my blender pipeline and export a T3D compatible model that has LODs and proper MIP maps, basically everything, in just a few dozen hours from 'I am a newbie to this photogametry thing'. ADDITIONALLY, you are correct, diffuse, normals, heck even ambient occulusion and be baked in. The most impressive thing was the integration into my 'newish' Blender pipeline. Where I can bake in 'edge mapping' and a ton of other things. I so need to put out a new video showing a comparison. I am thinking a 'reboot' of the conifer pack using using these new techniques versus the pack everyone knows.
In my experience so far, This IS NOT FAST, it's NOT easy and I am getting by because I have A LOT of experience with 3D. Even my 'big machine' 16 gigs ram, plus dual nvidia 780 ti GFX cards has been GRINDING away at some of these scans. Granted, I am a newbie at this and I see and understand things even better than I did last night. I am planning a video to explain it all but, it's not a 'beginner' task by any means.
More to follow.
Ron
#25
That was me, actually.
And yeah, there's still some manual work involved, but the end results are so detailed it's worth it.
If you want to know how well it scales to large things, check out The Astronauts' blog covering how they used the technique for the Vanishing of Ethan Carter.
They were able to scan entire buildings, rock quarry walls and even a mountain valley via helicopter. As long as you provide it good, consistent pictures, it can build the model.
The blog: www.theastronauts.com/2014/03/visual-revolution-vanishing-ethan-carter/
And here's an image from that blog showing a 3d model in their game they got via scanning:
@Ron
Yeah, I've been pretty impressed with it. Laser scan would definitely have it's place, particularly for visually indiscernible things like foliage or other depth-complex things, but for the majority, it produces some pretty incredible results.
We're definitely going to have to pow-wow on this and hammer out a nice, documented pipeline for this.
11/10/2014 (9:38 pm)
@Duion That was me, actually.
And yeah, there's still some manual work involved, but the end results are so detailed it's worth it.
If you want to know how well it scales to large things, check out The Astronauts' blog covering how they used the technique for the Vanishing of Ethan Carter.
They were able to scan entire buildings, rock quarry walls and even a mountain valley via helicopter. As long as you provide it good, consistent pictures, it can build the model.
The blog: www.theastronauts.com/2014/03/visual-revolution-vanishing-ethan-carter/
And here's an image from that blog showing a 3d model in their game they got via scanning:
@Ron
Yeah, I've been pretty impressed with it. Laser scan would definitely have it's place, particularly for visually indiscernible things like foliage or other depth-complex things, but for the majority, it produces some pretty incredible results.
We're definitely going to have to pow-wow on this and hammer out a nice, documented pipeline for this.
#26
But I found projecting regular photos of an object on a low poly version of it, works also, even if you have only one photo from the front, because you can reconstruct the rest with some tricks, I did this with my first weapon model: opengameart.org/content/stg44
You can also photograph an object from all angles, then projection paint the texture on it in the modeling program, but only works for simpler objects.
11/11/2014 (4:05 am)
The problem is now where does the average person get a helicopter and high quality cameras to do scans of buildings...But I found projecting regular photos of an object on a low poly version of it, works also, even if you have only one photo from the front, because you can reconstruct the rest with some tricks, I did this with my first weapon model: opengameart.org/content/stg44
You can also photograph an object from all angles, then projection paint the texture on it in the modeling program, but only works for simpler objects.
#27
Getting a high quality camera is easy. You can get 15-20 megapixel cameras for under $100.
And yeah, you can also manually model via the photos if that's the direction you choose, though you lose out on the high res model that the photogrammetry produces, so you'd have to normal map the textures, or do those manually, both of which wouldn't provide as nice of results.
There's several ways to go about this pipeline though, which is always cool.
11/11/2014 (5:49 am)
Well, obviously using a helicopter is extreme. But capturing most of a house isn't too hard if you can get a few high vantage points as well, such as from other buildings.Getting a high quality camera is easy. You can get 15-20 megapixel cameras for under $100.
And yeah, you can also manually model via the photos if that's the direction you choose, though you lose out on the high res model that the photogrammetry produces, so you'd have to normal map the textures, or do those manually, both of which wouldn't provide as nice of results.
There's several ways to go about this pipeline though, which is always cool.
#28
With elevated viewpoints it gets even more hard, in most cases you have no elevated viewpoints, if you are lucky you have one elevated viewpoint, but this is not enough, since I think you would need at least 4 elevated viewpoints from different positions.
Yes, the camera is not a big problem, but I would not use some ordinary camera for under $100, I thought so, too in the beginning, now I have a more professional one and the difference is like day and night. It was a lesson I learned, never compromise on equipment, you will have to spend more time later compensating lower quality source material, your time is more valuable.
11/11/2014 (6:34 am)
I already tried to capture buildings from all 4 sides, but already this was hard, because you cannot just step into peoples garden or in most cases objects are bocking the view in at least one of the directions you need.With elevated viewpoints it gets even more hard, in most cases you have no elevated viewpoints, if you are lucky you have one elevated viewpoint, but this is not enough, since I think you would need at least 4 elevated viewpoints from different positions.
Yes, the camera is not a big problem, but I would not use some ordinary camera for under $100, I thought so, too in the beginning, now I have a more professional one and the difference is like day and night. It was a lesson I learned, never compromise on equipment, you will have to spend more time later compensating lower quality source material, your time is more valuable.
#29
Scan#1: ref photo 1
.
SCAN #1;
22 photos using a 'cheapish' (130$) GE 400X 14mpx camera.
Scan Ref photo #1:
and SCAN #2;
20 photos using the same camera. (Oh yeah, its also UNDER 900 polys)
Total 'Modeling Time' 55 minutes to get the model and the texture maps.
Ron's got a NEW best friend.... :-) (does anyone else think the 3D looks way better than photos? Odd....)
I will be starting up a 'Photogametry and T3D blog shortly.' This MOST DEFINITELY warrants further research!
Oh yeah, real quick comment concerning helicopters and expensive camera rigs etc. Uhm, I did a bit of research (because I am a huge nerd) and I found that a 'cheapo' remote control helicopter with a cheapo 'hero' camera (set to take shots every few seconds) can do some amazing stuff..... NO, it's not going to be 'perfect' but, for games... hey good bye terrain editor.... Yet another way this tech can work in our favor.
Ron
11/11/2014 (11:14 am)
Well, I think it's getting to be time to move this to a new blog but first! It was a GLORIOUS, OVERCAST, and Generally cold and miserable day today here in Texas. PERFECT for taking scan shots.... (Yeah rather counter intuitive but, hey it works.) So , I grabbed my cheapo camera a bunch of AA batteries and headed out to a local park. To view the full HD images, right click and select 'View Image';Scan#1: ref photo 1
.SCAN #1;
22 photos using a 'cheapish' (130$) GE 400X 14mpx camera.
Scan Ref photo #1:
and SCAN #2;
20 photos using the same camera. (Oh yeah, its also UNDER 900 polys)
Total 'Modeling Time' 55 minutes to get the model and the texture maps.
Ron's got a NEW best friend.... :-) (does anyone else think the 3D looks way better than photos? Odd....)
I will be starting up a 'Photogametry and T3D blog shortly.' This MOST DEFINITELY warrants further research!
Oh yeah, real quick comment concerning helicopters and expensive camera rigs etc. Uhm, I did a bit of research (because I am a huge nerd) and I found that a 'cheapo' remote control helicopter with a cheapo 'hero' camera (set to take shots every few seconds) can do some amazing stuff..... NO, it's not going to be 'perfect' but, for games... hey good bye terrain editor.... Yet another way this tech can work in our favor.
Ron
#30
Very nice stuff there Ron.
While it naturally raises the compute time, more images help better define the shapes, so that's something the consider. 20's probably OK for a small rock, but if you're willing to eat the compute times, more images promise better results.
I'll have to see about getting my sister's professional shoot camera and see what I can get out of it this week.
11/11/2014 (12:57 pm)
Hah!Very nice stuff there Ron.
While it naturally raises the compute time, more images help better define the shapes, so that's something the consider. 20's probably OK for a small rock, but if you're willing to eat the compute times, more images promise better results.
I'll have to see about getting my sister's professional shoot camera and see what I can get out of it this week.
#31
Now, I plan on running higher resolution meshes JUST for edge detection and normal mapping really (these were 'generate dense cloud' on medium just speed up processing. Either way, the two examples here are game ready and in the end they look 100% better than what I could do by hand...
Ron
11/11/2014 (1:08 pm)
@Jeff, True more images = better final but, in the end I am building this stuff for a game engine, and realistically, even the less than 900 poly version looks better than I could have modeled by hand. Also, there are NO holes or missing spots. I think I nailed the number on these. (As a note, I did overlap around 70 to 80 percent on these versus the recommended 50%). Keep in mind too, I have 2 other sets of photos that 'blew up' when I tried to make the models. On the plus side, I was able to 'target' the bad images, remove them from the project and the results were usable (with liberal use of 'cap holes in mesh' tech.Now, I plan on running higher resolution meshes JUST for edge detection and normal mapping really (these were 'generate dense cloud' on medium just speed up processing. Either way, the two examples here are game ready and in the end they look 100% better than what I could do by hand...
Ron
#32
Ah, I thought those were the high settings. If those were only the medium, then that is very solid results.
Yeah, in my testing I also tried to be over-zealous on the overlap to mitigate problems filling in. Even if you figure some fiddling needing to happen to get good results from the shots, it's still far less a time investment than producing the same quality level by hand.
So had you thought of how to actually make use of those sorts of assets in a map yet?
I was imagining stuff like that being akin to decals. You trim the mesh to have a blendable base and then place it to conform to the terrain, so it looks like it blends in naturally.
11/11/2014 (1:28 pm)
@RonAh, I thought those were the high settings. If those were only the medium, then that is very solid results.
Yeah, in my testing I also tried to be over-zealous on the overlap to mitigate problems filling in. Even if you figure some fiddling needing to happen to get good results from the shots, it's still far less a time investment than producing the same quality level by hand.
So had you thought of how to actually make use of those sorts of assets in a map yet?
I was imagining stuff like that being akin to decals. You trim the mesh to have a blendable base and then place it to conform to the terrain, so it looks like it blends in naturally.
#33
I DO have a totally functional workflow developed. It's not that complicated. It involves Photoscan, blender, and Meshlab...with a smattering of XNormal as well (for certain procedures). What kills me about this, dude.... slap on a Rift and DAMN....if I could come up with something that makes you 'smell' the forest....you would TOTALLY think you were there.
I really do think this blog needs to move to it's own little 'section of the web'. I really think once people see what is possible, even with 'budget' equipment, this subject will take off. Heck, I am TOTALLY reinvigorated thanks to your suggestion to try this. Thanks man.
For those wondering here is a breakdown of the total investment:
Photoscan Software: 179$ for standard license
GE 400X 14 megapixel digital Camera (Walmart 130$)
16 Gig Memory Card for Camera - 20$
Blender - FREE
XNormal - FREE
GIMP - FREE
Meshlab - FREE
T3D...duh
total cost to implement this: Approximately 330$ US.
In comparison I also went the 'laser scan and SSL route'
Software: 450$
Laser: 350$
Projector (keep in mind it's too weak to do what I want) 750$
Results.... photoscanning WINS hands down.
Ron
11/11/2014 (1:41 pm)
Ahhh, I have already started a demo level/video. The concept is simple, Conifer Forest Pack compared to what I can do with a combo of photoscanning and traditional modeling. There will be more to follow on this of course.I DO have a totally functional workflow developed. It's not that complicated. It involves Photoscan, blender, and Meshlab...with a smattering of XNormal as well (for certain procedures). What kills me about this, dude.... slap on a Rift and DAMN....if I could come up with something that makes you 'smell' the forest....you would TOTALLY think you were there.
I really do think this blog needs to move to it's own little 'section of the web'. I really think once people see what is possible, even with 'budget' equipment, this subject will take off. Heck, I am TOTALLY reinvigorated thanks to your suggestion to try this. Thanks man.
For those wondering here is a breakdown of the total investment:
Photoscan Software: 179$ for standard license
GE 400X 14 megapixel digital Camera (Walmart 130$)
16 Gig Memory Card for Camera - 20$
Blender - FREE
XNormal - FREE
GIMP - FREE
Meshlab - FREE
T3D...duh
total cost to implement this: Approximately 330$ US.
In comparison I also went the 'laser scan and SSL route'
Software: 450$
Laser: 350$
Projector (keep in mind it's too weak to do what I want) 750$
Results.... photoscanning WINS hands down.
Ron
#34
I mostly meant the specific usage of like, the above rock models, but yeah, no doubt you're well into usage of it 'for realz'.
I'm glad someone else is able to hammer some immediate headway on this. I've got other things to work on at the same time, so my ability to R&D into this is slow, if steady(haven't had a chance to nab the license for Photoscan yet, though that should happen soon). Added bonus is it's easy to back and forth with someone here than it would be to try and hit up the VoEC dudes. I'm sure they'd like to talk, but we don't have any pipeline hurdles to overcome here ;)
I've got a few ideas I've still gotta test that could push it yet another step up, but it's not 100% ready for me to say if it's workable or not. I'll definitely be following up with that as R&D continues on my side.
Glad to see you've got some fire back in you :D
I'm really looking forward to what you put out with it.
11/11/2014 (2:06 pm)
Nice.I mostly meant the specific usage of like, the above rock models, but yeah, no doubt you're well into usage of it 'for realz'.
I'm glad someone else is able to hammer some immediate headway on this. I've got other things to work on at the same time, so my ability to R&D into this is slow, if steady(haven't had a chance to nab the license for Photoscan yet, though that should happen soon). Added bonus is it's easy to back and forth with someone here than it would be to try and hit up the VoEC dudes. I'm sure they'd like to talk, but we don't have any pipeline hurdles to overcome here ;)
I've got a few ideas I've still gotta test that could push it yet another step up, but it's not 100% ready for me to say if it's workable or not. I'll definitely be following up with that as R&D continues on my side.
Glad to see you've got some fire back in you :D
I'm really looking forward to what you put out with it.
#35
Ron
11/11/2014 (2:28 pm)
Well thanks man. I have been working and trying to get back on the 'tight edge' of this stuff and thanks to you and your efforts, I can firmly say I have some ideas that will probably put T3D ahead of the 'Next Gen' since I can honestly say this new gen of games does NOT look all that good. The best game I have played from a visual standpoint was Vanishing of Ethan Carter....and that is on an OLD UnReal 3 engine! Hell, I was a DIE hard CryEngine guy up to that game. Anyway, trust in the fact that I will make something TOTALLY amazing with these new techniques and T3D 3.6...1 .7.9. 4.001 ;-)Ron
#36
"The best game I have played from a visual standpoint was Vanishing of Ethan Carter....and that is on an OLD UnReal 3 engine!"
Wanna know why it looked so good?
Because of the blog I linked above. It looked as good as it did because they were using this exact process. ;)
11/11/2014 (3:02 pm)
@Ron:"The best game I have played from a visual standpoint was Vanishing of Ethan Carter....and that is on an OLD UnReal 3 engine!"
Wanna know why it looked so good?
Because of the blog I linked above. It looked as good as it did because they were using this exact process. ;)
#37
11/11/2014 (3:06 pm)
Can't wait to start using this, it looks like a lot of work to redo everything, but it may be worth it.
#38
.
I know it's IMPOSSIBLE to tell without a frame of reference but, it works. (Could be better but, hey its a start.)
@Duion, I highly recommend at least trying this. In fact, I am taking all the stuff that you are countering with and using that as a basis for my 'testing' so PLEASE keep thinking of all the stuff that makes you say 'hmmmm, this might not work in this case...' In fact, This 'cliff' test was based on what you said about needing 'high vantage points' From what I have researched and used, the program 'counters' the focal points and adjusts the image to ensure you have the right scale. (Freaky, I know)
I need a few folks to 'stress test the system' so to speak. So, THANK YOU and keep the ideas coming.
Ron
11/11/2014 (3:32 pm)
Just a quick additional note concerning 'height'. This image is an overhang I shot this morning to try and do something 'higher' than I was. The technique was a simple test. I shot the 'level' parts 6 photos. Then I took one step back and angled the camera up (trying to maintain the overlap). Keep in mind, I am still 'experimenting' with this whole thing. The resulting 'cliff' is actually about 25 feet tall. .I know it's IMPOSSIBLE to tell without a frame of reference but, it works. (Could be better but, hey its a start.)
@Duion, I highly recommend at least trying this. In fact, I am taking all the stuff that you are countering with and using that as a basis for my 'testing' so PLEASE keep thinking of all the stuff that makes you say 'hmmmm, this might not work in this case...' In fact, This 'cliff' test was based on what you said about needing 'high vantage points' From what I have researched and used, the program 'counters' the focal points and adjusts the image to ensure you have the right scale. (Freaky, I know)
I need a few folks to 'stress test the system' so to speak. So, THANK YOU and keep the ideas coming.
Ron
#39
EDIT: get it? rock? I crack myself up.
11/11/2014 (3:45 pm)
You guys rock! Can't wait to see some results from this. Like, I'm actually getting energised just seeing you guys being so excited.EDIT: get it? rock? I crack myself up.
Associate Ron Kapaun
3tdstudios.com
This is interesting, we need to talk via skype or whatever sometime.
Since I am making 'game models' I don't think I need all the 'fancy' measuring and calculating features. Let me get a trial copy of the standard license version and try it out.
Ron