Revolution scan

Prototype V2 goes online

Tue, 12 Nov 2019 13:25:00 +0000

Product Design

Prototype V2

As soon as I built the first version, there was a moment of realization. It is complicated, really complicated. It is far from a machine I would build again or have someone else do it. There were way too many different parts, some portions were flimsy and would even fall off. Therefore, I had to go back to the drawing board again and start from scratch. V2 was born, mostly printed. After all, if you can 3D print a 3D printer, why wouldn’t you be able to 3D print a 3D scanner.

I decided to pack everything neat and tight in the base of the device and to split it into 3 logical portions: base, body and top. All the electronics are to be placed in the base, and to have cameras and lights in the body and top parts.

Initial tests

Well, they turned out promising. Raspberry Pi cameras seem to be major pain and I had some success only when I placed a +2 macro lens on top of it.

Putting it altogether… once again!

After days of fiddling and set backs I finally managed to put it all together. There was a huge bin of failed parts, until I’ve reached the complete machine. Here it is next to the old one: smaller, better, slicker.

Lights on!

Let’s test it.

Dino to the test!

Click Here to build one on your own.

Building Prototype – Part 2

Sun, 07 Jul 2019 10:06:49 +0000

Product Design

Building a prototype – Part 2

This is where I try to improve the rig by building better lighting and using better lenses on the cameras.
I figured that better uniform lighting all around the rig, combined with decent lenses, might be the winning combination for the scanner. If my reasoning is correct, I might actually be done with this.

I started out by fixing equally spaced LED strips onto the flat surface, together with connections. It was quite a laborious process, but I hope that, if the lighting is done just right, scans will be flawless. Since there is a lot of work to be done I needed some help.

Lenses

This is the pain point. I spend hours and hours of reading different forums and different sites to get any information on what lenses would be the most suitable for this kind of task. I ended up choosing varifocal lenses, which I think would be my best bet in this case.

Putting it altogether… again!

Time to fit all the things back together. I started pretty much from zero and wrapped it all up in a closed off “case”. Yes, it’s ugly, but it feels much sturdier than previous attempts.

Lighting it all up with Lighting McQueen. My phone didn’t adjusted well to all the LED’s.

Let’s test it.

I have run several test scans on this rig and can tell the results are somewhat satisfactory. It’s not exactly what I was hoping for, but improvement, albeit slow and painful, is visible. Some new issues have emerged, such as heat. So many LED’s make it quite warm inside, so they might need cooling.
Aside from that, the rig is ugly. That’s the truth, and something to address in the next iteration.

Building Prototype – Part 1

Tue, 28 Nov 2017 21:07:47 +0000

Product Design

Building a prototype – Part 1

This is where it gets messy. I ordered electronics and parts and started building this thing up.
I’m not an engineer, but I do have some experience in electronics and soldering and what not so I expect this to be a bumpy but a rewarding road.
With some help from my engineer friend, we crafted a prototype using Raspberry Pi’s, Adafruit Motor hat, and a DC motor.

More about it bellow.

As I mentioned, solution consists of 4 Raspberry Pi Zero W boards, each with attached Pi V2 Camera on it, Adafruit DC motor HAT and Pololu DC Motor. Pi’s communicate between themselves through WiFi and are accessed through it as well. One Pi is assigned to be the Master and other 3 are Slaves. The reason I opted for this solution is to avoid mechanical challenges of lifting a heavy DSLR in 3 or 4 vertical positions. Instead this will be an elegant solution where only the subject rotates and cameras taking shots after each rotation. Main Pi will turn the platform and signal the other 3 to take the photo each time. They respond back and the next rotation of the subject takes place.
Altogether, it looks something like this:

Light it up!

I need some light in here. I made two rings out of LED’s and that alone took me a whole night of just soldering. I really have no clue how to construct this, so what I did is I cut the two discs and connected them with PVC pipes. Overall, the thing is stable, but I have to add an oscillator to keep the LED’s from warming up. Some parts have to be 3D printed, so I used the great service 3D Hubs provides, to have the custom parts for the rotating platform printed.

Cameras, Lights, Motors…

Testing the rotation platform. It’s alive!

Time to close it in and test, test, test.

This works… sort of.

There are a lot of issues with Raspberry Zeros and Raspberry cameras, though. Camera flat cable connector on the R Pi side is pure crap. It breaks so easily, I managed to mess all four of them. Had to hot glue them back in place. I don’t know what they were thinking placing those I have some cheap Chinese flat cable converters, and those open and close just fine, no matter how many times you plug and unplug them. Another big issue is, those cameras cannot focus on a near subject. Doesn’t matter what you do to them, it just doesn’t work. See the pictures bellow. I’m ordering some additional lenses and I will try with them once they arrive. I suspect the cheap pinhole lenses just don’t work. Will see how that goes.

Proof Of Concept

Fri, 29 Sep 2017 10:06:25 +0000

Product Design

Proving I was right (or not?)

I decided to go ahead and test it to see if my idea really works. There is no point in building anything unless we know whether it is going to work or not.
I reached several conclusions during this test:
1. I was right! It does work, but…
2. … it’s more complicated than I anticipated.
3. I have no clue how to build it.
Read on to find out how the testing went.

Initial test consisted of one Lazy Susan from Ikea, one cardboard box, a couple of large sheets of white paper, and some LED strips. I made marks on the platform to divide it into equal rotation steps and moved the camera up and down arbitrarily. Yes, it’s still a manual method, but hey, the point is to prove the concept.
Initial results failed multiple times until I realized something. Everything is bright white and the reconstruction software has no clue what is where or where the floor is.
That was when Melman got green screened.

Well… this worked!

And it worked rather well. Once I removed the background from it and ran it through PhotoScan, results were more than decent. Having a good mask changes everything. I used the one generated from the green screen removal and boom, everything aligned perfectly.

Green screening worked to an extent.

Issue is, there is a lot of green spill over the colors as you can see bellow, making it look dull, and messing up the colors.

To fix it, I changed the approach. I left the white background, used the auto masking feature and added some markers to tell the program where the floor is. Once done, result are quite amazing.

In conclusion…

I found what works and what doesn’t work. So, let’s discard what doesn’t work and focus on doing more of what does work. Sounds simple enough.
Not really. Building an actual device is a completely different beast.

Moments of Inspiration

Wed, 17 Aug 2016 13:42:23 +0000

Product Design

First ideas for A 3d scanner DRAFT

The “there must be a better way” was ringing in my head for a while, so, as soon as I got some free time on my hands, I sat down to try and conceptualize it.
First couple of ideas were just crap, but hey, how do you come up with a good idea if you don’t have any bad ones in the first place? Anyways, this is what I’ve learned from extensive brain storming:
1. I need a rig similar to full body capture scanning rigs.
2. I need as few cameras as possible.
3. Camera must not rotate around the subject.

Going back to the original plan, this has to be as affordable as possible, preferably under $1000 per unit to build on your own.

My initial idea was to have a subject centered in the middle of this weird contraption that will move the camera in the front and the background plate behind the model, spinning all around it and taking photos at each step. While conceptualizing it, I soon realized this solution would never work for several reasons:

First, the whole thing would be enormous. If you add the model size, plus distance to the camera, plus the camera size, you easily get to something that is at least 80-100cm in all directions. That is just way too big.

Second, it would be mechanically “challenged”. The number of rails and moving a DSLR around is just not practical. It would cause all kinds of malfunctions, besides beinh heavy and impractical to move around.

Finally, camera would block the light. This would cause shadows falling all over the subject.

In conclusion, camera has to be fixed in front of the subject that would rotate around itself.

Concept 2 : A (somewhat better solution)

Going back to the drawing board (modeling software in this case) and spending a couple of sleepless nights, brought about this brain fart. Camera arm is supposed to pull out from the base which contains all the electronics, including the motor for the base rotation, rails attach to it, with the camera sliding up and down the system.
I was quite happy with this, but my gut feeling told me that this wouldn’t really work.
Why?
Well, it’s not automated enough, for one. While triggering can be solved via remote IR LED, you’d still need to pull the camera up and down on your own, and I have no clue where would one store the rails while not in use.
It made for a good base for tinkering, though.

In conclusion…

Well, I haven’t succeed in finding a good solution.
I haven’t failed either.
The whole conceptualizing process was done on a computer and while I haven’t made anything I came to some conclusions running the simulations, renders and modeling which, I believe, will lead me to something decent in the end.

Humble Beginnings

Mon, 18 Jul 2016 08:33:43 +0000

Product Design

Humble beginnings in the world of 3D scanners

First of all, I’m not an engineer. I’m a CG artist. Second of all, being a modeler too, I tried getting my hands on a decent 3D scanner everyday use that is:

1. Affordable, costing under $1000.
2. Gives good results usable in production.
3. Can recreate textures along with scanned mesh.

I wasn’t able to find one. Thing is, good ones are way too expensive. I would have to use it day in, day out to even come close to justifying its price. Cheap ones are mostly laser based, so the resolution is not acceptable. Worst of all, they do not scan the color. There is no acceptable solution that’s under $1000 which gives out good models that will require minimum of post processing work.
So, I decided to build one. As i said, I’m not an engineer, I used to do programming a bit in school before I decided to become a CG artist, so I expected this to be a long and a difficult challenge.

This all started with an assignment I had 3 years ago. We were doing McDonald’s TV ad that, of course, featured a burger and fries. Food had to look delicious, even as a 3D model. I approached the assets modeling by dwelling on 3D scanning an actual burger. Food was ordered, lights were set and the meticulous process of photographing it from all possible angles started. I spent good half a day taking photos, and the second half processing the images and building the dense clouds and fiddling with settings. Results were acceptable but required a lot of manual work to get the production ready assets. This got me thinking, there’s got to be a better way!
(It also made me hungry).

There’s got a be a better way!

This has haunted me ever since. I looked everywhere, but could not find a solution that would suite me. Problem is, I do not need a scanner on a day to day basis, and the price of “professional” solutions is just prohibitively high.
Products on the lower end are some laser based scanners, mostly open source. However, their output quality is just not good enough for a decent production use.
I also don’t want to setup the lights and stage every time I need to make a scan of something, and most definitely I do not want to walk around with a camera and figure out myself if I covered all the angles, and whether there are enough photos to make it work. Furthermore, I want repeatable results, so if something changes I want to be able to scan it again in a month or so, and still get the same results, with same quality, and same lighting.
Therefore, there’s got a be a better way!

I just have to find one.

Since my quest to find an automated scanning solution for under $1000 miserably failed, I figured there is only one way to do it. I’m going to build one my self. That is one bold statement. As I said earlier, I am not an engineer, nor a product designer, but I do have some technical education in electronics, and some programming skills, even though I haven’t used those in a while. Surely I can make 3D models, so, I guess I stand some chance to make something decent and usable. We’ll see.