- Interview with Dr Adrian Bowyer, the creator of the RepRap 3D Printer
- The Mathematics of 3D Printing
- 11 Reasons why Schools need 3D Printers
- 3D Printing a 3D Printer
- How to change the battery on an IP54 digital caliper
- Off Topic – not 3D Printing Related
Photofly is a recent offering from Autodesk to enable you to create a 3D model from a series of photographs.
You can get it here. On the page it says “Photo Scene Editor” but this is it’s Other Name.
Yes, this piece of software has two names, which seem to be interchangeable. Confusing? Yes. Not as confusing as my first effort with the software!
EDIT AFTER PUBLICATION: Brian from Autdoesk writes:
By the way: Project Photofly refers to the web service (in the cloud) that does the computation (a lot of math running on a lot of CPU’s and GPU’s), while the “Photo Scene Editor” (PSE) is the name of the Windows desktop applicaiton that you work with your models and with Photofly. Photo Scene Editor edits photo-scenes and Photofly creates photo-scenes.
And another thing – it’s been emotional
When I was doing this review there was a lot of frustration encountered with the software. This is the first 3D modelling software I’ve used so I am not sure of some of the terms used. I am a complete noob in that respect. I think that in my noobiness made the job a touch harder than it should be. Full Disclosure for you!
Installation was easy
Installation was simply a matter of downloading the file and installing it. It handled the fact I’m on a 64 bit Windows 7 without a drama.
Really, I didnt expect anything less. You need to give them your email to register the software but this was not a burden – it is in beta so far.
Taking Pics of The First Item
I think software should be intuitive and natural in it’s use so I’m not one to read a manual.
So I started off with a toy that I had around, placed it on a mug and took a photo, twisted it a few degrees, then took another photo etc. I was careful to ensure every side of the image was photoed. This is the resulting photo set:
So with that done, I fired up the program.
The program looks like this. It seems quite simple.
Usually you click on “Create a New Photo Scene”
Then it will ask you to upload a your photos.
If you are having trouble selecting photos from your computer then you probably shouldn’t be driving a computer.
We send it to the Cloud
After you have selected you images the software uploads them to Autodesk’s central server.
This is a really key point. All the uploads and calculations are done in the cloud. So don’t upload any images that you want to keep private – particularly those papparazzi shots you got the other day!
The advantage of the cloud is it allows Autodesk to release a free thin client, while keeping their (I assume proprietary) algorithms nice and safe on their internal server. Is this a good business model? My thoughts are, probably.
Wanna wait or should we Email You?
By far the best (and perhaps most innovative) aspect of the software is after the images are uploaded, a lot of calcs need to be performed. This takes time.
The software will ask you after you select your images if you would like to wait for it to process, or if you would prefer an email from them when they are done.
I tried both methods over time and it worked fine.
By the way, every time you do stuff with the finished model you need to reupload and recalc so it is worth resizing your images to a reasonable file size. I found 800×600 about right.
Mkay – model has been calculated – lets check out the UI
Once your model is finished calculating you will get an email, or the progress bars will complete and you will be presented with the main UI of the software.
Fair dinkum – it’s looks messer than my dog’s breakfast. Other than having a vaguely convex shape – it doesnt look anything like the figure.
Well that is because I didn’t read the manual!
Turns out reading the Manual will save you time – who knew?
Yeah ok, I’ll admit it, I was wrr. I was wrrrrr. I was wro…. Nup can’t say it.
A few things I learnt after the first 2 failed models (and 2 hours of creating new swear words) is that:
- The light and Contrast on the Object Matters.
- You have to walk around the object to photograph, not keep the camera in one place.
In the photo guidelines it makes these two points quite clearly. There is even a video tutorial of some dude walking around an incan mask and photographing it from all angles.
Anyway, I tried again.
Here is a Bucket what I photographed:
Finally some success!
As you can see the model created is vaguely bucket like.
But what is that gap in the side you ask? Well that is the contrast from the buckets causing a shadow on itself and hence the software can’t cope.
EDIT AFTER PUBLICATION: Brian from Autodesk writes:
You are correct that the bucket is not a good object to use. The reason is simpler than “subsurface scattering”. The reason has to do with two key factors:
(1) smooth plastic, polished surfaces, or glass are often “shiny” meaning that the algorithm gets confused by “seeing” the _refelections_ of the background off the shiny surface. For a given point on an object the reflection color changes based on camera angle. Even if your object isn’t shiny, if the surface you place it on is shiny it can ruin the process.
(2) The bucket’s surface has no surface detail. It is monochromatic. Every point on the bucket’s surface looks like every other point. Photofly works by matching visual “features” (a speck of dust, a freckle, a scratch, etc.) between all the photos. When there is nothing to match there is no way to triangulate a position on.
But this was my first feeling of getting somewhere. Total time take? About 3 hours.
Ok now I get this photo thing
I then went out an looked for subject that were preferably in shade (for consistent contrast with no “sunny bits”) and ones that I could walk around.
I found a stump near my backyard that I took a few photos of, uploaded , calculated and waited.
Here is a sample shot,
And here is the model in Photofly.
Now we are really getting somewhere! This model looks stumpier than my mate’s little finger!
I started to get a bit of a rush at this point – we were almost there!
Ok – time to talk about the interface.
Twisting and turning the model is quite simple – it is all mouse driven and depends on where the mouse is in comparison to the model as to whether you pitch, roll or yaw. It is very straightforward to pick up.
Hovering over each item in the menu gives you enough of a description that you know what you are doing. Look Dad – no manual!
A Stitch in Time
It turns out that even though the Autodesk calculation software is pretty good, it is not perfect. You still need to tell it how each photo you have taken relate to each other.
You do this by marking similiar points between photos.
This is called stitching. Here is a sample I’ve prepared earlier:
While I’ve been a bit dramatic about how much time it took to take the shots and process them, once you have the Knack, that is the easy part.
THIS SCREEN is where you will spend most of you time. Yes, stitiching the photos together is where you will get squinty eyed and a hunched back marking the points.
The process is straightforward. You can see in the photo the green circles – these are points that I have marked as the same on each photo.
Ideally you need 4 points as green.
The red square points are where the software has spit at me and said that they are not the same point. Well I said it was and submitted it anyway.
Once you have gone through EVERY photo you have taken and marked the points ( the software gets about 50% of them right between shots) you are ready to resubmit and get you final model.
The more photos you take the better the model will be
After you have stiched all the photos, increase the mesh quality (top left corner) submit again and then export to Youtube.
Here is the final stump in all it’s glory with awesome fly over:
The video cannot be shown at the moment. Please try again later.
And to prove it wasn’t a fluke, here is a Statue in my garden (the jitter in the photo was me ducking behind a tree so I couldn’t get an even degrees of photos):
The video cannot be shown at the moment. Please try again later.
So in the end we got it working!
Theorectically we can export the model to 3D Print later. For the Technical – cause I know you want to know: I was able to export the file as an OBJ file and then import it into their other product Autodesk 123D (which is a CAD program) and then I could save as an STL from there.
In Conclusion – here are my tips:
- Select a photo with a consistent light ambiance (ie no sides in direct sunlight causing the other sides to be black)
- Take heaps of photos
- Move around the object to take shots
- Stitching is your friend.
My final rating on the software is 4/5.
It is easy to use once you understand what sort of photos to take. The rating could be higher is there was an STL export function and if the stitching was a bit more automated.
You can get Photofly here.