Showing posts with label 123D Catch. Show all posts
Showing posts with label 123D Catch. Show all posts

Saturday, November 23, 2013

Exploring the Sense 3D Scanner

A lot of people think I work for 3D Systems.  Not only do I NOT work for 3D Systems, I am forever surprised to find that I'm late coming to the party when they release something new.

This was the case of the Sense 3D Scanner.

I had not idea they were even contemplating a low cost 3D scanner... especially one that retails at $399.00!  My first clue came from an email sent to me by a friend that pointed to an article on the web announcing it.  That article, in turn, pointed to the page for the Sense.

I was floored!!!  Talk about being out of the loop!  LOL!

But, I quickly made sure that I had one in house to test.  It arrived late last week and I quickly ran through a few tests.

Before I talk about my tests, I have to put into context my expectations.  First, I cannot afford a $30,000 3D scanner.  So, I have no idea how the Sense stacks up to the big kid toys.  The only "scanning" with which I have any experience is using 2D to 3D applications like 123D Capture.   I blogged about this technology in 2012.

My benchmark for a low-cost 3D scanner is that it has to be (1) easier than 2D to 3D techniques and (2) has to produce equal or better results.  If it meets those two goals it will be very useful to me.

My First Scan

Obviously, when one takes delivery of a 3D scanner, one has to find a suitable target to scan.  I found mine in a character from Monster's Inc.  I didn't go around the object a full 360 degrees because I simply wanted to see if it got ANYTHING.

I also didn't fully understand the options as I went through the workflow.  So, I inadvertently removed some detail.  I was amazed at how easy the process was and how well it came into Sculpt.

I wenr on to scan a styling manikin head that I'd previously used in 2D to 3D tests.  Here is the result as seen in Sculpt.  Because I wanted to print this without supports, I added a bottom and back to create a bookend affect.  The object then prints laying on it's back.  It's printing as I write this.  At the size that I am printing it will take 10 hours.  I'll take a photo when it's done and add it to the bottom of this post.

When compared to the process for 2D to 3D, this was infinitely easier.  And, in terms of being able to handle things like hair, it was amazingly more effective.  Here is a short video of a complete head scan.

Now, don't get me wrong.  Some things work and others do not... like very small white sculptures.  And, to be able to scan reliably every time, is going to take some practice.  But, the feedback that we get as we scan is very helpful and the software's capabilities to create a solid object from a partial scan are nothing short of miraculous.

Here is a short video that demonstrates the basic workflow from scanning to saving an STL file.

The software developers have done a wonderful job of making an intuitive interface that provides excellent feedback during the entire process.  The is made it easy to capture and edit something the first time it was used.  But, reliable results across a wide range of subjects is going to take a bit of experience.  In the little time that I've had it, I have been able to learn as much from the failures as I have from the successes.  All white objects, for instance, might require some special lighting to increase shadows, etc.

In all, I probably only have 2 hours of experimentation into using the Sense 3D scanner.  But, initial observations are mixed when it comes to using a tripod mounted scanner pointed at an object on a turntable.  But, that might be because I have a manual turntable and my hands are in the field of view.  I've also learned that it is best to put the object in an open area when scanning via a turntable.  Otherwise, close fixed objects seem to confuse the scanner.  Rest assured, I will put a LOT of time into coming up with the most effective strategies for using the SENSE.  It's a fantastic device at a fantastic price!

Sunday, September 8, 2013

Sculpt Tutorial #5 - Repairing Poor STL Files

There are a number of developing ways to create 3D objects based on already existing items.  Scanning, while expensive now, is destined to both drop in price and yield more useful results.  Another way that several companies, including 3D Systems, are pursuing is turning a series of 2D images into a 3D object that can then be printed on a 3D printer.

But, all of these techniques can have difficulty capturing things like hair or highly reflective surfaces.  This is where an application like Cubify Sculpt can be invaluable.  As a test of these capabilities, I decided to see what I could do with an object that I created several years ago using the beta version of 123D Catch (Then called Photofly).

The item captured was a hair styling manikin.   I sometimes use these to test new photo lighting configurations. 

Styling Manikin Head

123D Catch did a great job on the facial features.  But, the hair didn't turn out as well.  This is common in 2D to 3D capturing.  The defects in the hair made for a poor 3D print.  Until Cubify Sculpt was released I had no way to correct the problems.  I've been waiting for an application that could fix the head so that it printed well.  So, it was natural to bring it in to Cubify Sculpt and see what we could do with it.

Here is a video that explains the process and shows the outcome.

While we do not need to paint objects that are to be printed with the Cube 3D printer, I decided to go ahead and see what I could to color the head.  The painting functionality of Sculpt is pretty basic.  So, it turned out to be the most difficult part of the process.  My primary criticism has to do with the inability to make the bush small enough.  And, my primary suggestion would be that the developers add the ability for us to place an image, for each axis, in the background to help select colors and refine shapes.  What does help is to use the BLENDING function to create the appearance of finer features.  

It was a lot of fun tackling this one.  And, it encourages me to look into Cubify Capture to see if it will allow me to capture members of my family to create 3D printed sculptures.  There wasn't much point for me to attempt to use Cubify Capture before this because I know that it would probably have the same issues with hair, etc.  But, now that I know what Cubify Sculpt can do to repair that problem it just might be time to give Capture a try!

Tuesday, May 15, 2012

2D to 3D Print - Sparky the Fire Dog

I thought it would be helpful to show a second 3D print that originated as a series of standard photographic images.  This time I printed out Sparky the Fire Dog.

Sparky is a large remote control vehicle that fire departments use at events to help them teach children fire safety.  Here is one of the images that I took at the Montgomery County Fair (Maryland).

Sparky The Fire Dog R/C Vehicle

There are some interesting things about this subject.  For instance, aside from Sparky, himself, most of the features are relatively shallow.  Also, the crowd passing by means that the background scene changes, at least to some extent, in every image.  I was pleased that the 123D Catch engine seemed to be able to sort this out and find points with which to stitch the images together.  Finally, the light on the top of the car is semi-transparent and that DID give the 123D Catch engine a bit of trouble.

Here is the video that shows the resulting 3D object rotating in space of my first attempt to capture Sparky.  Only 6 images were successfully stitched in my first attempt.

Fortunately, I was able to go back to the fair and take another series of images.  This is the result of the second attempt.

As you can see, this one was MUCH better than the first.  And, it is this "Catch" that I used in my 3D Printer test.  Notice, however, that the light on the top still gave the engine some problems, creating some holes in the mesh.

Here is the final print done on my RapMan 3.2.  The Cube would produce a smoother outcome.  But, this is good enough to show you that it is possible to create and print a 3D model of real objects without having to draw a thing!

Sparky The Fire Dog - 3D Print
Considering the shallow features, the issues raised by the semi-transparent lights and the issues my current 3D printer has with the "Y" axis slipping, the result is quite remarkable.  And, it can only get better as the 123D Catch engine is improved, my expertise with clean-up improves and I get to use the Cube to create the final print.

Monday, May 14, 2012

From Point & Shoot Camera to Cube Printer

Some time ago, long before I knew about the Cube, I became interested in a program by Autodesk that purported to turn a series of 2D images into a 3D object.  When I began using it, the program was called PhotoFly.  They later folded the PhotoFly project into their 123D initiative, which is a suite of 3D programs designed to be easy for users.  It's now called 123D Catch.  Right now it can be downloaded as a beta.

In experiments, while the technology was a lot of fun, it turned out that the mesh that it created wasn't absolutely clean.  So, there was some question as to how useful it would be for the real application that I'd hoped to use it for... 3D printing.

Last week, the topic came up in the Bits From Bytes forums and I decided to give 123D Catch an opportunity to see if I could come up with a way to use it with a 3D printer.  This post is about that experiment. 

First, let me show you 123D Catch in action and explain how it works.  It starts with a series of photographs like this one...

And then we bring all of the photographs into 123D Catch.

123D Catch Application
At the bottom of the 123D Catch application you see the series of photos.  At the risk of further confirming to my neighbors that the guy next door is more than a little nuts, I set up a hairstyling manikin head on the rail of the porch of the back yard studio and shot a series of images while walking around the head.  (I know what you are thinking.  But, it's all innocent enough.  I used the head to perfect studio lighting techniques many years ago. LOL!)

Bringing the series of images into 123D Catch, the program calculated the position of the camera for each shot and then stitched together the images to create a 3D model.  From the front it looks like it did an excellent job.  But, that is only partly right.  As seen from this rotated view.

Holes in the 123D Catch Mesh
The problem is that 123D Catch has issues with dealing with hair.  It cannot seem to find stitching points because hair just does not give it enough differentiation.  This is a problem.  Here is a side view that shows the confusion.

The hair has confused 123D Catch so much that it maps grass onto the head!  Obviously, this is NOT going to do all that well in a 3D printer.  A 3D Printer needs a "Watertight" mesh to print well.

Fortunately. we have a solution.  We export the 123D Object into a format that can be read by a wonderful free program called NetFabb Studio Basic.  NetFabb's job is to find and fix problems with mesh and to export a clean STL file for printing.  When we bring it in, the offending holes are clearly visible.

NetFabb on Entry

A huge warning sign tells us that we have a problem that needs to be fixed.  Clicking on the "+" sign brings up a dialog that allows us some options for fixing the mesh.  Here is the resultant fix.

NetFabb Fix
As you can see, the hole in the back of the head has been filled.  It's not perfect.  But, at least it allows us to export an STL that we can print.  And, that is pretty remarkable considering how big those holes were!

So, how did it print?  Take a look.

Original Photo
RapMan 3.2 Print of 123D Catch Object
Frankly, I was skeptical so I first printed it very small to see if it was even usable.  I was astounded by the fact that it so closely resembled the original!  So, I printed it out much bigger and was even more impressed.

Frankly, I'm astounded by the result.  It even got the hair flip!  The nose probably could be a bit better.  But, who's complaining.

What this means is that we will be able to create 3D prints on our Cube 3D printers of our loved ones!  I plan to capture my grandchildren this weekend.  Because the 123D Catch engine has issues with hair, I will try to add something that gives the engine something to lock onto like a colorful hat or ribbons.  This is no small deal and it is capabilities like this that will go a long way to winning over the 3D printing skeptics!

I'm certainly a believer.  :)

NOTE:  I expect the Cube print to be noticeably cleaner than what you see in these images.