649 réponses à ce sujet

[Tarot Redhand]

and try to find some mathematical pattern in the whole thing.

@MD - Chaos theory?

 

TR

[MerricksDad]

Got the pattern, just need to tweak the smoothing/optimize combo. It comes across in the game a little harsh on the seams, and there are a few tiles with some heavy noise marks too close to each other.

 

[Tarot Redhand]

The above would be awesome as a snowy tileset with very little adjustment (I had to look twice before realising the white is just untextured!).

 

TR

[MerricksDad]

 

[rjshae]

Surprisingly, it bears a certain resemblance to my unmade bed in the morning... it was a "rocky" sleep.

[T0r0]

That is a thing of beauty !!!!

[MerricksDad]

My wife said the picture with the rocks poking out was like the oreo creme filling with chocolate stuff, like brownies and chocolate bars. The longer ones are like fudge bars.

[3RavensMore]

I can hardly wait to see that textured.  --swoons--

[3RavensMore]

Thinking out loud...  Imagining an automated system for building a whole integrated area landscape.  Something that reads a user created bitmap layout of terrain types / altitudes (think something akin to topo map) and creates all of the tiles and walkmeshes (based on what was outlined above) as one single integrated user defined area.  The system would then slice the whole thing into tiles, exports everything into one unique tileset, and build the .set file.  The builder could populate it with trees and other placeables at will, but not alter the geometry of the area in the toolset as all the tiles will only fit together in one way.

 

I know, impossible, right?  I'll admit I have only a vague idea of the limits of altitude changes, and how pathfinding is set up in tilesets--the most I've done is copying existing tiles/groups and adding geometry within those tiles. 

 

Edit: While I'm dreaming, perhaps it would be possible to define certain areas of the bitmap that would allow tileset based structure subsets to be placed.

[MerricksDad]

I started this journey over 10 years ago. Back when I was first working on my recreation of the bioware infinity engine. I had named it the "dragon engine" and it was able to take in topo map coordinates, and render (without the help of a true 3D engine) Sims2 style height tiles. My original Dark's Gate campaign was created from maps of the black hills region around Keystone, SD.

 

Nothing about what you said was impossible, because I know for a fact it has already been done. It is (and was) very simple to convert everything I had made for the Dark's Gate module (mark 1) over to the tileset I created in 2008, using GFF editor, and custom programs to modify height coordinates in the area file. All I had to do was set the grid scaling, look into the topo map data (which was downloadable from USGS sites before google earth was popular), and read a series of numbers from the raw data in intervals related to the scale I picked. For the dragon engine, I once rendered a map in infinity engine miles (to the character unit scale), but unfortunately my custom render method took 2 days to render just the ground plane (pixel by pixel rather than polygon shapes).

 

Today we could easily take that same data, or perhaps the data I played with for the canyons on Mars, and create actual NWN maps from portions of that. And with these scripts I am working on now, with a processor stronger than mine, which could do the whole thing from import to NWN file export in a single pass, the whole thing could be automated:

 

1. Import USGS data
2. define scale
3. detect largest z transition over a single tile
4. detect proper height transition value
5. dynamically create a tileset (without plants/rocks) to represent the base ground
6. select a region from the USGS data
7. create an area file from the selected region
8. open and finish populating by hand

No it would not be that hard. I had already fully scripted about 10 of the hard coded walkpath nodes into my 2008 gmax scripts. It would be slightly harder to code for all of them (I think), but if I can get this far, it is not impossible to go the rest of the way. I have however come across certain transitions, especially ledge-like areas, where none of the NWN walkpath nodes will work, and I have had to fudge tile groups, lying about the actual pathing. This makes multi tile diagonal maps very hard to create for players who point and click once to move. For those moving constantly to a point in front of the mouse pointer, it works just fine.

[Tarot Redhand]

or perhaps the data I played with for the canyons on Mars

 

MD, does that mean you have a martian tileset laying around, that you haven't told anyone about?

 

TR

[MerricksDad]

Nope But believe me, I have already collected mars textures for just that reason. I can't get enough of the pictures of mars, where the rocks suddenly go from that weird metal gray, to the rust colored sand. I love that color transition.

 

 

 

[Tarot Redhand]

To my untrained eye, the darker grey rocks in the top picture, look like very large lumps of pumice. I hadn't noticed that before. Hope the skies are clear enough for everyone to see the (super) lunar eclipse in a few hours.

 

TR

[MerricksDad]

I'm sure bubbly molten rock flying though space has some appearance similarities with pumice. The other meteoric pictures I have are shots from cross sections housed in the SD School of Mines' geology museum. If you get a sec, I suggest walking through that floor of the building. Got some dinos up the middle and US mega-fauna from the time of woolly mammoths and saber-tooth tigers are around the edge, including, I think, an Aurox skull, some camel horses, a cat bear, and some pig dogs. Back wall has a giant carp looking fish head I could park my car in, which isn't saying much since I drive a Chevy Spark...but still, impressive. Many of the minerals on the one side of the hall are all from local mines.

[MerricksDad]

I don't know if anybody can imagine this as vividly as I do, I have always imagined that at the time olympus mons blew its last big load, there was a giant frozen ocean of carbon dioxide and water on the surface, or just under the first few layers of rock. When parts of it then melted, creating the massive washouts from the top of the planet to the southern sea, the frozen in-place metallic pumice sunk to the sea floor as the water was evaporating, or going to hide further underground, while the carbon dioxide evaporated and was then deposited back on the glacial poles. It would be interesting to find out geologically how accurate that is. But that is just what I get from viewing the global pictures. Especially the "chaos regions".

[MerricksDad]

I spent about 15 minutes tonight working on some weight loss programs for the higher poly models. I reduced them to about 2/3 what I have shown above by simply relaxing them a pinch, smoothing them with a 15 threshold, then optimizing heavily keeping smoothing groups. They now export super fast, and retain almost all their detail. I guess that method is kinda like optimizing GIF images with an octree, but not as labor intensive to write code for. Now I just made the mistake of re-smoothing them before I exported, but otherwise they look and feel great, as before. I just need to have the similar script chop, texture, and run a few fixes, and we're good to put trees on it, happy ones or otherwise.

[3RavensMore]

Hey, a Bob Ross reference.  Love those happy trees. 

[MerricksDad]

Hey, a Bob Ross reference.  Love those happy trees. 

As many as possible with me

[MerricksDad]

So after getting no good diagnosis yet on the leg and bruising issues, I decided to review everything I had done in the last month, excluding the first bruise I had taken, in case it was not related. I had jogged to the bus with the kid one day, which could explain the impact issues on the shins, and I had spent a lot of time sitting in a very uncomfortable chair, which pushes on the veins in the backs of my legs. I tested a few theories while I wait, but all I found was that exercise through movement didn't fix the issue, but staying off the veins in the back of my legs for long periods of time did help drain out whatever was ailing my legs. I got to looking up Gout, because OTR mentioned it, and I thought about getting tested directly for that. I have read that it can be brought on by beans, and I remembered having a TON of beans the week before my legs pretty much shut down.

 

Last night I had reached a point of healthy feeling that was similar to over a month ago. So today I decided to test a theory. I made another batch of White Chicken Chili, with the same beans, minus the green onion. 2 hours later I am feeling a tingling tiredness in my lower legs, and my leg that bothered me the most during the last month is starting to ache again. I may have found at least the key to finding out exactly what is going on. Now to have myself tested for uric acid buildup.

 

In the meantime, I will probably keep eating the chili until it is gone, but just not overdo it, and make sure to hydrate and keep myself moving until it stops again.

[Tarot Redhand]

and make sure to hydrate and keep myself moving until it stops again.

Errr, isn't that counter productive? Shouldn't you keep the exercise routine going even after it stops as a preventative measure.

 

TR

[MerricksDad]

I don't intend to stop moving, at least not until I can get my brain placed into a robot I just mean, at least keep active this week, instead of being mostly at the computer like a few weeks ago.

[MerricksDad]

Wrote a little script this morning which examines all the edges on a mesh and determines which ones could benefit from being flipped. This basically makes the terrain more smooth, and sets up the mesh to behave better with softening and optimizing techniques.

 

This seems like it took forever, because first I had to determine if the edge indices changed after being flipped, which they don't (not exactly), and then I had to learn that even if a function reports that there is X edges, a trimesh actually has X*2 edges in reality, because faces don't technically share edges. Then I had to write scripts to figure out the edge values AFTER an edge was flipped, by first determining which verts they were going to flip to AFTER being flipped. Bah.

 

Anyway, if anybody wants the edge smooth function for gmax, I can share it on here, or you can wait until I update my library on the vault. I've got quite a few recent mesh modifying additions to put out.

 

Spoiler

/*
	OPTIMIZE EDGES
	
	Usage: optimizeEdges obj
		where obj is the object node you wish to modify
		*may not function properly if the stack is not clean
*/

fn arrayPush a x = (
	l=a.count
	a[(l+1)]=x
	return a
)

fn arrayIndexOf a e = (
	if (a.count<1) then return -1
	i=1
	while (i<=a.count) do (
		if (a[i]==e) then return i
		i+=1
	)
	return -1
)

fn getPolyVertsNotSharingEdge obj edge = (
	revEdge = meshop.getEdgesReverseEdge obj.mesh edge
	faceList = meshop.getFacesUsingEdge obj.mesh (#{edge}+revEdge)
	faceVerts = meshop.getVertsUsingFace obj.mesh faceList
	edgeVerts = meshop.getVertsUsingEdge obj.mesh edge
	r = (faceVerts-edgeVerts)
)

fn getEdgesOnlyUsingVerts obj vertList = (
	edgeList = meshop.getEdgesUsingVert obj.mesh vertList
	--now subtract all the edges using other verts
	edgeList2 = meshop.getEdgesUsingVert obj.mesh -vertList
	r = (edgeList-edgeList2)
)

fn softenEdge obj edge_index = (
	verts = meshop.getVertsUsingEdge obj.mesh edge_index
	newEdges = getEdgesOnlyUsingVerts obj verts
	edgeArray = newEdges as array
	outerVerts = getPolyVertsNotSharingEdge obj edgeArray[1]
	faceList = meshop.getFacesUsingEdge obj.mesh newEdges
	faceArray = faceList as array
	v1 = getFaceNormal obj.mesh faceArray[1]
	v2 = getFaceNormal obj.mesh faceArray[2]	
	a1 = acos(dot (normalize v1) (normalize v2))
	meshop.turnEdge obj.mesh edgeArray[1]
	
	modEdges = getEdgesOnlyUsingVerts obj outerVerts
	modEdgesArray = modEdges as array
	faceList2 = meshop.getFacesUsingEdge obj.mesh modEdges
	faceArray2 = faceList2 as array
	v3 = getFaceNormal obj.mesh faceArray2[1]
	v4 = getFaceNormal obj.mesh faceArray2[2]	
	a2 = acos(dot (normalize v3) (normalize v4))
	if (a1<=a2) then (meshop.turnEdge obj.mesh modEdgesArray[1])
	
	OK
)

fn getEdgesAsVertPairs obj edgeList = (
	vertPairs = #()
	edges = edgeList as array
	for e = 1 to edges.count do (
		edge = edges[e]
		--check for twin edge
		revEdge = ((meshop.getEdgesReverseEdge obj.mesh edge) as array)[1]
		revEdgeIndex = arrayIndexOf edges revEdge
		--if this is not the reverse edge add it to the list
		if (revEdgeIndex > edge) then (
			arrayPush vertPairs (meshop.getVertsUsingEdge obj.mesh edges[e])
		)
	)
	vertPairs
)

fn optimizeEdges obj = (
	--store all edges as vert pairs
	vertPairs = getEdgesAsVertPairs obj #{1..obj.mesh.edges.count}
	--for each vert pair get the edge
	for p = 1 to vertPairs.count do (
		e = getEdgesOnlyUsingVerts obj.mesh vertPairs[p]
		e = (e as array)
		if (e.count==2) then (
			edge = e[1]
			softenEdge obj edge
		)
	)
	OK
)

 

It has one known flaw, and that is if you wanted a sharp angle, it will try to remove it. If that angle is surrounded by multiple other sharp angles, it will make a mess, and can flip a face. Multiple uses on the same object will often make such a mistake even worse.

 

I still need time to document everything, but there are already too many projects.

 

For instance, instead of running this on all the 600 tiles, I need to go take care of a city flower bed I run.

 

Edit:

 

I don't know if this is just Gmax, but this line is crashing the gmax script runner with a "system error" very often.

faceList = meshop.getFacesUsingEdge obj.mesh (#{edge}+revEdge)

If you move the compilation section to a new line and instead have code like this it has less errors.

edgeList = revEdge+(#{edge})
faceList = meshop.getFacesUsingEdge obj.mesh edgeList

That is a system error without any details, and then if I try to do something else, including closing gmax, it just crashes the program. Something about how it is handling the inline merger of two bitarrays is having issues. I'd like to see the code for gmax, but I guess that won't likely happen any time soon.

[T0r0]

Flipped edges ???   

[MerricksDad]

Let me show you with a picture or two.

 

Sorry that image is visually painful. Here is another one to lessen your pain, or for the color blind.

 

 

 

In the first image, I have highlighted a section where faces fold over each other. I could not possibly hope to create a walkmesh from this mesh because face edges reside over others in 3d space. But since all the verts are separate from each other in XY position, I could flip the edges, un-tucking those faces, and then apply a walkmesh without issue. The faces in pink are the faces of issue, and all adjacent faces. I have smoothed this by 45 degrees so you an see the offending faces are darker and unsmoothed.

 

In the second image, I have run this script one time. It unfolded the majority, but the pink area shows one tucked face, surrounded by adjacent problem area. Again, I have smoothed this by 45 degrees to highlight the tucked face, but since it is fully under another, you can only see that an edge line passes through the space of other faces. Still can't make a walkmesh from this mesh.

 

In the third image, the final offending edge was turned and the area is finally free from walkmesh error. I have smoothed it again by 45 degrees to show that the entire region is now smooth, and that no faces are tucked under another.

 

This all happens without moving a single vert. Now compare the face edges from mesh one and mesh three, and you can see other regions of the mesh were also smoothed better when other edges were turned.

[T0r0]

Ahh ok. Thought there was something else I had been overlooking all this time (phew ! ). This is still about the faces tho. Albeit your way , without moving verts around is very cool. Does it actually highlight the bad faces or was that for illustration purposes ? Does it work in 3dsmax ?