Seeking a blender artist for a small 3D Modelling job

Hello. We are Cruncher, a web development agency.

We have a model of an oak tree that has been reconstructed from a scan of an actual 300 year-old tree. The model is large: it currently has 70,000 faces. But it is full of holes where the scanner has not been able to ‘see’ parts of the trunk and branches. You can see this in the attached screenshots. We are looking for someone to spend some time patching up the holes. Many of the holes will be patchable almost automatically, but some are large and will need some artistic judgement about how to reconstruct the form of the tree. The final model should be a closed solid with normals pointing out.

There is no texturing or lighting required for this task: it is purely modelling. The final result will be rendered in WebGL and is part of an interactive visualisation that will be available online soon.

The deadline is the 8th April, in one week.

If you are interested and available, please PM me with your estimate (in Euros, Pounds, Dollars, CHF, whatever) and an example or two of your modelling.

Stephen Band
Cruncher

Attachments



Actually it seems much easier to to model a tree that overlays the scanned tree then to fill all the missing holes, because that would take much more time then to model an overlay tree with decent topology

Remodelling has also been suggested in a couple of private replies, with a suggestion it would only take a few hours. Here’s a bit more information.

The point of this model is not to get the lowest polygon count possible (although that would also be nice), it’s to get an accurate representation of the real-world, existing tree. The current 30,000 vertex model is a reduction of a scan of 2 million vertices. My concern about remodelling is that we end up too simple, and we lose the detail of the original tree.

The model is to be used to tag the branches with scientific data taken from the real tree, so it has to be a pretty close copy.

I can’t see a model of this complexity only taking a few hours: there is lots of detail in all those branches. Simply experimenting with the data reduction algorithms took me a day to get a semi-acceptable result (I used a combination of Blender and MeshLab).

So I’m unconvinced about remodelling - but I could be wrong.

Retopology would keep all of the details.
That is, by using retopology, subdivision and shrink-wrap on the mesh a modeller can get the most optimal results.
It’s best if you sent the 2 mil polygon model to the modeller and he’ll do the job.

Thanks for your input!

The original 2 million vertex scan is not a model, it’s just a cloud of points. And it’s noisy, so I had to do some clean-up using MeshLab algorithms before converting it to a model.

I attach a couple of screenshots of the original point cloud to give some idea of what that would be like to work with.


Attachments


That looks like a lot of work :smiley:
And the point cloud looks very unclean… a good mesh just can’t be generated from that…
Hmmm… Better send the modeller all you have and he/she will figure it out.

Freemind has a good point, if your 70k mesh was generated from that point cloud, even then you cannot be so sure if it is really accurate since the the generated 70k mesh probably left details out that are in the point cloud, my best bet is to take the point cloud and model the tree based on the point cloud with subdivions of course so it will look smooth. it might take a bit longer then a few hours i think a day or one and a half would be sufficient. if you like my proposed method i would be willing to take this little job, however i am not going to fill thousands of holes, because that is kind of lame, sorry to say. :slight_smile:

My Modeling works you can see in my blender videos

See one Tree and Root Animation work here

Looking for freelance Blender work. See on of my Tree Animation Free Model here

Hello all.

Thank you to everyone who responded to this thread, both privately and publicly, and for the discussion about the best approach. We’re going with the remodelling approach, because as @rogerbootsma said, the alternative is lame.

We loved the quality of everyone’s work, and we’re moving forward with one of the first to get in touch. Seems only fair. I’ll post a link to the final result when it goes online in a few weeks, in case you’re interested.

Cheers!
Stephen