40 camera headscanner, results

waterwingz

12138

Re: 40 camera headscanner, results

« Reply #10 on: 05 / December / 2017, 19:33:10 »

Advertisements

Quote from: ottokar on 05 / December / 2017, 18:08:26

If you help someone start a similar business in another city, you could charge them for your help, but once they have a working setup, they wouldn't feel the need to pay you an ongoing fee, they could just stop paying you and carry on using the hardware and the scripts that you've set them up with.

The trick here might be to "rent" out the necessary SD cards. With a little customization of CHDK, there are several things you could do to protect your business. Things like making the SD cards hard to copy and also ensuring that they only worked for the agreed "rental" period. At some point, it gets to be easier to pay you than try to hack around the safe guards.

Logged

Ported : A1200 SD940 G10 Powershot N G16

mphx

210

Re: 40 camera headscanner, results

« Reply #11 on: 19 / December / 2017, 18:44:34 »

@ottokar

Lets see your questions

1) We bought projectors because we were dissatisfied with the quality from Agisoft Photoscan.It has like a zillion pro features , but lacks a lot in the quality department and speed too.
We knew that cameras couldn't do any better..we know we had to do so much manual processing to get a better model..but as it turns out RealityCapture software can produce a very high quality model even without projecting pattern .Of course since we have the projectors we use them for even better results.The difference in quality is chaotic.The difference in speed chaotic.
So if we knew RC earlier or better if RC was out earlier , we might had not buy the projectors.Its an additional cost.And not a low one.

2) Even with projecting pattern to models..the end-model , needs work.It is in a very good level but it needs few touches here and there.We can make a model ready for printing in 2 hours (of manual work)..so we are ok with that.So yes we dont need more cameras or better cameras as far as quality is concerned.
But we'd like to have more cameras because when we have to shoot kids cameras dont have good coverage.The "high" cameras miss the kid entirely.So we have to move , to turn down some of the high cameras to get a better "view" of low models..kids...animals.With more cameras we could avoid this extra "job" of turning up and down cameras when we have to shoot animals and kids.

3) 3 projectors in 360/3 angle positions cover every angle (projectors have wide fov ) ..we'd like some higher stands though , since we use low stands to put them on and we miss some spots.
We dont really care about the "back" projector if he loses spots especially high in head.Its hair , we will fix them manually..noone cares

We care about the 2 "front" projectors to cover correctly heads..faces...thats the most important spot for us.Its hard to be reconstructed if it isn't modelled well.
So if projectors cover body and head (front view) we are ok..if some tip of the head or some feet (ankle and down) is missed...we really dont care..we will recreate it manually afterwards...

4) Our shooting setup is a two-man procedure.It can be done with one person but when u have the client , children , animals running all around , its better to be done with 2 persons.
One computer is controlling the cameras , one low spec laptop/pc is controlling the projectors.Projectors are connected to each other at a cascade way (projector1-projector2-projector3-laptop/pc) so every projector is showing what the laptop's screen is showing..that is ..the pattern

One person handles computer , chdkptp , shooting commands etc.The second person handles laptop and the switching of patterns (pattern-blank-pattern-blank etc)
Shooting command is a 2 shoot procedure.For technical reasons first shooting is the patterned one , second one is the normal.
So the procedure goes something like this.Man2 turns on pattern.Man1 issues the shooting command.Man2 hears the first click so switches off pattern.Second clicking is heard.End of shooting.
2 shoots are like 5 secs away.We tested various times , we couldnt go lower than 5 secs..cameras were losing sync for various reasons.So we stayed at 5 secs , i believe every adult human being can stay relatively still for 5 secs..small movements because of lets say breathing is no problem for us.
Animals and kids , are single shots.Many single shots ,some with pattern and some without and we do some cut and paste..head from one..body from another...texture from a third one..worst case scenario

They were many businessmen that wanted to start a similar/franchise business like ours.To give them instructions/know-how/guides on hardware and stuff.
Our perspective on that matter is simple.You will get all these on a nice one-time fee of some thousands euros and a contract that you will never (well not never, for some XX years) open a similar business in the same town we are .Go open it in the next town , in another country or in the Moon.We dont really care.Just not in our town.
Because if you do , its a move that says "we want to get a bite in your pie/clients".Simple as that.
If someone is willing to pay our one-time fee , sure we will help him.
We dont need to get paid in ongoing basis..we give them everything at a nice price.We will even buy the hardware for them..and then do whatever you want.As far as the above rules are met.

« Last Edit: 19 / December / 2017, 18:48:45 by mphx »

Logged

ottokar

29

Re: 40 camera headscanner, results

« Reply #12 on: 27 / December / 2017, 14:57:41 »

Hi Waterwingz, mphx

Quote

The trick here might be to "rent" out the necessary SD cards. With a little customization of CHDK, there are several things you could do to protect your business.

My initial thoughts were that because the scripts are in plain text, and the sd-card *has to be* readable, then it can't be hidden, it has to be in plain sight, and perhaps instead one would have to learn how to recompile CHDK, and then compile / somehow incorporate the script directly into the CHDK binary files - so that it is not readable in a text editor, and furthermore, make CHDK check the camera serial number before booting up, if possible.

- if one goes down this route, it's definitely yet another thing to think about and get one's head around. Mphx's method of just having a large one-off fee would also be another way of doing it.

Mphx - so much information again. I'm sure photogrammetry, and using CHDK to do it cheaply, will take off massively in the future, so your answers will benefit a lot of people.

I haven't yet done any re-modelling of the 3d scans - my first impression was that it would be too labour intensive, but if your experience is that 2 hours work is enough to get clean models, that doesn't sound too bad.

It's good to hear from your practical experience what is important and what can be skipped over, and about needing 2 people just to handle the customers, things like that.

Quote

We tested various times , we couldnt go lower than 5 secs..cameras were losing sync for various reasons.

I imagine the 5 seconds is a lot to do with the amount of time it takes to write each photo to the sd-card? If so, then what I had in mind, if I use projectors, was - if you have 64 cameras, then dedicate 40 for projection (and building the model), and 24 for texturing, so that you don't have to wait for a camera to finish writing to the sd-card before doing the 2nd photo. This would allow a shorter interval than 5 seconds - have you also thought about this and did you reject it for some reason?

It's very interesting that a lot of businessmen wanted to do a similar business - I presume they have no special connection to 3d stuff, but their interest is purely that they see it as something new that customers would pay money for?

andrew

Logged

mphx

210

Re: 40 camera headscanner, results

« Reply #13 on: 31 / December / 2017, 11:33:31 »

Hey ottokar ,

I will try to combine some of your questions in my following things i will write.

First of all , 64 cameras are not enough , for so many reasons.One of them is the coverage of adults and kids i mentioned in a previous post.
Another reason is the low spec cameras.That's why everyone goes for 100+ cameras or dslrs.
With e.g. 100 dslrs you take a near-perfect model , and you just send it to the printer.Almost zero work is required.
But with compacts , you need post-shooting work .2 Hours is the minimum for someone that knows what must be done , knows about modelling , sculpting , painting.

So what "businessmen" want to do since they don't know anything about 3d ? Easy to setup system , many cameras , cheap and near-perfect result so they don't need to rely to some 3d designer to "fix" the model.They just want to send it directly to printer.Easy money.Well , there is no cheap system to do that.You need , as i said , many dslrs to come close to this scenario.And dslrs are not cheap.

So , we are all stucked with crappy compacts and ofc you can't get enough of them , because cost starts to go high.You take few , you fix the model fast and "profit".For 2hours work + 30mins of shooting minus printing cost = a nice amount of money

As far as what you suggest about splitting cameras for projection and texturing it can't be done.As i said above , cameras are not enough as they are.You can get a decent model with only 40 (we did it once , rest of cameras turned off and we didn't notice

) and get texture from 24 carefully selected cameras...sure it might be possible.
But i am sure that these 2 hours will increase , more work would be necessary , since less cameras = less details = more imperfections.
There must be a balance between profit (price-expenses) and the work that must be done for producing the end result/model.If we waste a whole "working day" to "ready" a model...its not profitable anymore.

And in periods (summer , christmas , easter , etc) were the clients are increased , you cant make one model per day...

Dont forget that in Greece , there is noone that prints.There was one , he stopped.Another one , his prints are crappy as hell.We send them abroad and we have a "delivery time" that we can't control that we have to calculate in the whole process.

EDIT : We couldn't go below 5 secs.We tried alot of things.We did some benchmarking on cameras , we tried to use same kind of sd cards in all of them.Below 5 secs wasn't reliable.
Either sd cards don't perform the same?Lag because of mechanisms in cameras?Chdk to blame?Combination of all ?I dont know.
Anyway , adults can stand still for 5 secs.kids and animals are whole different story.We take single shots..either patterned or/and non-patterned and we combine body parts..usually a good face with a body..depends on parents what they will choose.
Animals...dogs are more cooperative to be honest.Just keep a treat in front of them and they kill keep perfectly still for hours

« Last Edit: 31 / December / 2017, 11:37:54 by mphx »

Logged

ottokar

29

Re: 40 camera headscanner, results

« Reply #14 on: 13 / February / 2018, 11:02:30 »

Thank-you mphx, very interesting reading

Logged