Archived post by have_you_tried_google.

tough one to get the exact shapes without more control and art direction, but a start….

little closer

Attachments in this post:
http://fx-td.com/houdiniandchill/wp-content/uploads/discord/20232311/28/23/squashMetalVellum_HIM_v002.hiplc
http://fx-td.com/houdiniandchill/wp-content/uploads/discord/20232311/28/23/squashMetalVellum_HIM_v005.hiplc
http://fx-td.com/houdiniandchill/wp-content/uploads/discord/20232311/28/23/image.png

Archived post by fabriciochamon

I’ve been fiddling with the onnx sop to get a clue on how to feed the data, since it’s not super intuitive at first sight. Here’s a commented hip if someone is interested. (style transfer / resnet image classifier)

Attachments in this post:
http://fx-td.com/houdiniandchill/wp-content/uploads/discord/20230911/23/23/onnx_examples.hiplc
http://fx-td.com/houdiniandchill/wp-content/uploads/discord/20230911/23/23/image.png

Archived post by esttri

huhu, yes its late but luckily i am a night owl :D.

if you want to fetch your rig with a component script you will need to use those 2 nodes. because you want your graph to be loaded as a graph objects. its a looooot nicer to work with

and here is a little goodie. a hello world component. add as many hello world nodes as you like 🙂

Attachments in this post:
http://fx-td.com/houdiniandchill/wp-content/uploads/discord/20230611/22/23/hellowworld_component.bgeo

Archived post by midgaard78

With regard to the color space talk, I happened to have closely investigated what H20 is doing just yesterday. The issue I was running into is that Karma’s output respects the Input File Rules set up in the OCIO editor (Edit > OCIO Settings…). So even if your working space is set to ACEScg, the rendered image will still be linear rec.709 unless you take some step to prevent it.
– You can put the string “ACEScg” in the render’s filename, and Karma will obey the file rule. – You can change the file rule for *.exr to ACEScg. If you’re already pre-processing your inputs, this is what you should do. This is the “correct” ACES workflow. – You can set the Output Colorspace parameter for each AOV that should be ACEScg. – Or you can use the OCIO filter in the render settings to apply a post conversion. I recommend against that one, though, as I do not believe the conversion is lossless, and it risks a double conversion.
The SDR 1.0 Video view transform from ACES 1.3 ought to be sufficiently visually similar to Output – Display – sRGB from ACES 1.2, but I haven’t verified that yet. The Un-tone-mapped view should be the same as Input – Texture – sRGB.
The ACES 1.2 config can be found here: github.com/colour-science/OpenColorIO-Configs/tree/master/aces_1.2

Archived post by technically_artist

Posting this for future reference and so others won’t walk my foolish path “`c // Vertex Wrangle vector quad[] = {{0,0,0}, {0,1,0}, {1,1,0}, {1,0,0}}; vector tri[] = {{0,0,0}, {1,0,0}, {0,1,0}};
int vtx = vertexprimindex(0, i@vtxnum); int vtxcount = primvertexcount(0, i@primnum);
if(vtxcount == 3) v@__intrinsic_uv = tri[vtx]; else if(vtxcount == 4) v@__intrinsic_uv = quad[vtx]; else // Probably should do NGons at some point v@__intrinsic_uv = -1.0; “`