jordibares

Hi Rich... maybe Medium? Weebly? Patreon? maybe even Blogger?

Rich, just to let you know Tumblr puts a massive data sharing wall in your website and I would dare to suggest you move your stuff somewhere else. Many of us won't be signing to Tumblr due to the privacy concerns and massive data gathering for user profiling they do. Anyway, just though you may want to know. Cheers

What about implementing this in Houdini? would be great! http://stanford.edu/~jingweih/papers/quadriflow/quadriflow.pdf

Very Generous of you, thanks so much for sharing.

IMHO SESI needs to invest on both areas (and others as well). Houdini needs a great modeller and a great render engine that fully takes advantage of the architecture in Houdini. None of those render engines do.

Houdini getting easier, more accessible is a matter of time, so those that feel the need to protect their jobs I would suggest to look at both ends of the spectrum, either you bring some truly serious maths and engineering skills to the party or you bring some serious good flair and taste. Don't fight gravity... you will loose.

Seems like you have to tweak the weights so they don't penetrate each other...

vector3 is just the type of object where you store your data that you need to retrieve. a common method would be something like; hou.parm().eval() I hope this is what you are after http://www.sidefx.com/docs/houdini/hom/hou/parm_.html or as a tuple http://www.sidefx.com/docs/houdini/hom/hou/parmTuple_.html

Point Cloud tools optimised for large datasets (400 million points upwards) - I have a 2.2 billion points dataset so take a guess the time for of the PointCloudISO - Filtering (outlier removal, Feature estimation, registration) - Normal estimation - Point cloud sampling (RMLS) - Region segmentation (detectin of planes like floor, walls, etc..) - Model fitting may be?

I have seen a couple of workflows but most notably, authoring in MD is a no brainer, then do all the sim in Houdini so you have an integrated simulation pipeline where cloth responds to RBDs, etc... The normal approach is to sim on a simpler mesh and then drive the high res with it, this also allows for a much faster iteration if you have constant changes like we do in commercials... I hope it helps jb

Even better, have this one which goes through the whole thing. mediatum.ub.tum.de/download/800632/800632.pdf

Hi Nick, I am experimenting also with large point clouds datasets but find there is a ton of work to do to manipulate these in an effective way and finding the right algorithms to estimate normals, filter outliers, classify semantically the point cloud, etc... The method you are seems to me like a dead end as the pointCloudISO SOP is very slow so large datasets don't seem to me viable using that method. You can try adding a normal (0,1,0) should do, and see how it works... Regarding meshing those, I feel this requires specialised tools given the amount of points... Even pcopen struggles... I guess needs some extra amount of effort... have a look at the things Ia m looking at... Good luck

Agree, GridMarkets service is really really good and their plugin, IMHO, is great.

It is really really good, I wish the wind did affect the trunk to be able to judge the animation though... as it is the lack of movement in the main trunk is very distracting but I can tell it is going to look fantastic once finished.

It is called Three point perspective and you can find how it is done here https://www.math.utah.edu/~treiberg/Perspect/Perspect.htm and I think this is also useful