Pictomod3D. Create models from photos.

styxar

Greetings all. I am here pimping software to create models rather than models themselves. Apologies if this is considered poor taste... Here's the rundown.

Pictomod3D Version 1.0 is now available for download from styxar.com

This software is simple to use and allows you to create 3D models from photographes or images.

Click here to visit the Pictomod3D site and download the free demo.
(Click here to visit the Pictomod3D alternative download site.)




Quickly create textured models from scratch, increase the number of 3D models you can fabricate. Integrates easily into your mesh sculpting process. Pictomod3D is perfect for rapid prototyping or finished 3D models.

With this tool you can create a library of new objects or build new last minute requirements fast. Much simpler and more efficient than starting a fresh model from nothing. Use the ready to go meshes and textures directly or as great prototypes to kickstart the creative process.

Fast, affordable and simple to learn, the free demo download comes with a quick-start tutorial to get you up and running in minutes.

Setting up a project takes minimal effort, little more than selecting the background mask, painting in a little extra where needed, and then your 3D model is ready to output.

Use either images of a rotated object or even a few images from arbitrary angles to create textured 3D meshes in Wavefront OBJ or DirectX X file format. Includes a utility to easily extracted images from AVI and MPEG movies.




Create high or low polygon models, up to 128 000 polygons per model.




[Edit: Here's the raptor again, this time with 50 000 polygons.]

[Edit: And a cow, no texture, 20 000 polygons]




For more information visit styxar.com

Click here to visit the Pictomod3D site and download the free demo.
(Click here to visit the Pictomod3D alternative download site.)

Hope you find the software interesting and useful.

Adios,
styxar.

qubism

this is scary

flaagan

while the textures and polygon flow are both pretty bad, this is a rather neat method of quickly mocking something up.. depending on how powerful the software is, it could be useful for creating high poly meshes..
kinda interesting to see the old method of making a *real* model and then scanning it in take a new twist.

KDR_11k

I'm thinking "this'd be nice with a topology brush".

styxar

The in-built polygon reduction is designed to leave as much flexibility in reposing the model as possible, for the purposes of animation and such. Once you are satisfied with your animations or new pose, further polygon reduction can be applied.

The textures may or may not be suitable depending on the use of the model. However they are useful maps on which to paint new textures. I have left the output textures as they are in the above images so people can see what they have to work with.

So for instance, if the model was meant for a game, you would perhaps output it at 10000 polygons, modify it as needed for the game, polygon reduce it with traditional algorithms to 2000 polygons and then repaint the existing texture at higher resolution, its already mapped to the mesh.

Alternatively, if you just need a miscellaneous object quickly for a computer animation, you could use a model directly.

Daz

Would be interesting to try it out with some of these:

http://www.anything3d.com/model-portfolio_session/index.php3

edit: guess that wouldn't work too well since there is perspective in these shots so their feet aren't perfectly flat on the ground.

styxar

Although the primary, and simplest, method of using Pictomod3D is from level rotated shots of an object, it has a second method where you can use images taken from arbitrary angles. It requires more work to set up a project like this, and the output model depends on both the consistency between images and also how well they are aligned, good results can still be attained.

In fact, if the images are of symmetrical objects this extra information allows better models from less images. Even if the images are not ideal, you can get some interesting models.

KDR_11k

I doubt it compensates for deformation (since that'd require an advanced AI or something damn close to recognize the object and where it'll usually bend) so it's not likely the output would be usable without serious reworking. Animating that raptor would be a nightmare since he desn't have any definition of the moving shapes (e.g. leg structure, especially the hip). That's why I was suggesting the topology brush, you can take the volume of the scanned model and give it your own meshflow.

CheapAlert

I wonder how well it would work on sprites

I should give that recent US hostage dude a try :P

KDR_11k

Throwing a sequence of rendered images at the system produces a mesh that's nowhere close to the original, most notably the arms are missing.

Cheap: Forget it. I doubt it'd work on cartoons, either because those are low on color information and therefore recognizeable shapes. And they never match up.

styxar

For the serious 3D modeller Pictomod3D is primarily of use as a rapid prototyping tool, outputing models you can then modify in other packages. A topology brush, as a form of further polygon reduction, would certainly be a good way to get the final model exactly the way you wanted it.

The original raptor model that image is based on is only 1.5 inches high and itself has no hips either, so the model is fairly accurate.

The polygon reduction system used seeks to avoid producing long, thin triangles that are common to other methods, so that all the triangles are roughly the same size and shape, making arbitrary bending and stretching in external software easier and won't distort other parts of the model. Once you have the final shape and animation you want, producing long, thin triangle in further external polygon reduction is not a problem.

CheapAlert

Is there a fill tool for mask painting?

styxar

Yes, there are tools to paint both masks and anti-masks on either an image by image basis or all the images at once.

[Edit: Sorry, missed the "fill" part. You can paint using one of three brush sizes, and set certain ranges of color as masks.]

Regarding making models from cartoons or sprites, the images need to be of the subject in the same pose. So if you had a front view, side view and angle view where all the lines between eye level, hand level etc match up then you can make a model.

styxar

Here's the raptor with 50 000 polygons.
[Edit: Moved image to first post.]

Moz

I guess it would be good for creating references that are extremely accurate to their real world counter parts.

KDR_11k

Moz: You mean like EA's "proprietary face scan technology"? From what Daz said even that doesn't seem to give usable results.

styxar

Have added a cow mesh up top as well, 20 000 polygons and no texture this time, just for comparison. This was made from one of those standard Farm Toy collections of little plastic models. So if you needed a farm for a game or scene, you could quickly generate a whole zoo of animals.

ElysiumGX

I tried this program with terrible results, but I didn't spend much time on the masking. And it didn't like the files I had for sequences, so I had to go back and rename all 32 of them properly. Could you have it where the first file doesn't have to be something00.jpg? Instead just be a number, and begin with the lowest? I'll try again later with something more basic. This would probably work well with Silo's topology brush to work out the final results. It sucks that I'm only getting a 14 day trial on a program I wouldn't dare buy in its current state.

styxar

The program is doing something normally requiring laser scanning or the like, so for rotated images it does require a decent mask to work. The help files goes into detail about masking. If you have a clear background color setting the mask is fairly simple, you usually only have to paint out a few bits here and there. On the other hand, if the background color is unclear you will need to manually apply the masks.

You may find the tutorial in the help file useful as a demonstration of this. Following the tutorial, and a quick glance through the pictures in the help file, should set you up for better scanning.

With rotated images, some care has to be taken with the consistency across images too. The center of rotation should not change from image to image (the center of rotation is not necessarily the center of the model). Also an even speed of rotation is required. If you are rotating the models by hand you should be aware of this.

You can build a model from just 8 images (actually less, but 8 or 16 is practical) where you rotate the images yourself. Make sure to keep the center of rotation the same and move it in 8 (or 16) even steps.

gauss

for the stated purpose--that is, rapid prototyping, this does look like a handy piece of software. good work.