ObjReader - load .obj files at runtime

Good afternoon Eric,

I am having problems when using your examples (WWWExample) on an Android device.
It can import .obj files from external sources (http://...), but not from the device it self.
I have tried to import this asset from the project directory in Android ("jar:file:// + Application.persistentDataPath + /name.obj), but neither it does work (I have already access a sqlite database this way).

What could the problem be?

Thank you,
Inigo

 

I believe you need to manage permissions for Android apps; this topic might help.

--Eric

 

Hi Eric,

Quick question for you:
I'm trying to load the following OBJ file: http://keech.cs.washington.edu:8000/03001627/db549976c0b7b960d42ec7e303174a87/model.obj

Unfortunately, when it loads, I'm seeing a strange artifact (attached), as if the per-vertex diffuse colors are being set incorrectly -- at least that's my guess.

The OBJ seems fine, as it loads properly in MeshLab, other 3rd party OBJ Loaders for Unity. In fact, when I load it as an asset into a unity project, it seems fine too! However, it may be possible the OBJ format is not compatible with your plugin.

Any suggestions on what may be causing this? I intend to look into the source code, but figured I'd ask you first.

Thanks in advance for your support -- and for creating this awesome plugin! Cheers.

 

Actually the normals in the file seem to be messed up, as seen in Blender:



If they're recalculated and the file re-exported, the object looks fine.

--Eric

 

Hi Eric,

i just bought your ObjReader,
Now i have a problem with my models, some have more then 4 points under a face like:

f 11/94/11 64/87/64 63/67/63 19/56/19 23/58/23 20/57/20 12/95/12

your code just drop these and some parts of my model are missing.
Is there a way to implement something, so these parts are also in my model?
Or do i have to recalculate them?

 

Models should be made from triangles or quads; n-gons aren't supported, and ObjReader prints a warning about what's happening. You can use an app like Blender to triangulate faces, then re-export.

--Eric

 

Little problem with transparency:

Objects in 3dsmax having alpha = 0.75
Objects from Obj reader in unity having alpha = 0.25 (= 1 - alpha)
The reason can be read here.

So my Objects *.mtl has:

newmtl Knot
Ns 10.0000
Ni 1.5000
d 0.2500
Tr 0.7500
Tf 0.2500 0.2500 0.2500
illum 2
Ka 0.0000 0.0000 0.0000
Kd 0.5882 0.5882 0.5882
Ks 0.0000 0.0000 0.0000
Ke 0.0000 0.0000 0.0000
map_Ka textures\ground010.jpg
map_Kd textures\ground010.jpg

As it looks, your reader takes "Tr" instead of "d".

Could you offer a switch for selecting "d" vs "Tr"?

 

In the ParseMTL function, after dVal is assigned, add this:

Code (csharp):

1. if (lineStart == "tr") {
2.     dVal = 1.0f - dVal;
3. }

--Eric

 

I deleted the dll, imported theObjReader.cs and added those lines.
It works in Editor, but how to get it running on Android/iOS?

 

It works as-is on iOS and Android; there's no platform-specific code. The dll and source code are identical.

--Eric

 

Right, it seems I have a differnt issue.
The project that worked yesterday, doesn´t work today an Android anymore.
Maybe I have problems using persistant datapath....

The obj comes from inside a zipfile that gets downloaded and unzipped to persistant datapath right now.
Maybe I have to clean Cache and stuff before?

 

On Android you might have to mess with the manifest file.

--Eric

 

Ok, could you give me a hint how to do that?

(The interesting part is, that my project with ObjReader worked on Friday on Android, iOS and WebGL.
Since yesterday I don´t get it running on Android and WebGL (iOS not tested yet), using exactly the same project...)

 

Probably something else then, since that wouldn't have anything to do with WebGL. I'd recommend reverting the project to a working state and proceeding from there.

--Eric

 

Besides this:
I opened a new empty project in Unity, imported your package.
Importing file (pig, car ) in Editor works, but also not on Android.
Also tried another phone. Its so strange!

 

I've been told by other ObjReader users that editing the Android manifest works, but I haven't done that myself, so unfortunately I can't offer more specific advice.

--Eric

 

Hi Eric,

This looks perfect for a project I'm planning, can you tell me if it's possible to scan the contents of a directory and import any obj's that gets added to it?

Thanks
Will

 

That's definitely possible, although you'd need to do the scanning directory part yourself using the System.IO classes. Then you would use the appropriate path(s) that you get with ObjReader.

--Eric

 

@Eric5h5 What is the conversion between Ks, Ka, Kd, Ns and the respective Unity parameters you use for the .mtl file? My imported materials look overly dull after import...

 

Ka values are set to the material shader's _Emission property, Kd sets _Color, Ks sets _SpecColor, and Ns sets _Shininess. If the appropriate shader properties aren't present, then the values don't get set. It does occur to me that the standard shader uses different property names, so I should add code to check for those too. In the meantime, you'd need to stick with legacy shaders such as Legacy/Specular.

--Eric

 

@Eric5h5 But do you convert them 1:1? I mean Ka, Ks, Kd do have a range from 0 to 1 and Ns has a range from 0 to 1000. Whereas I think _Color, _SpecColor and _Emission range from 0 to 255, and _Shininess ranges from 0 to 1. That means do you multiply it just by 255 or divide it by 1000 respectively? I thought the reason that the materials look not shiny enough was because of a conversion error. So I just wanted to understand how those properties are converted? I mean shouldn't it be a non-linear conversion?

 

No, _Color etc. are float4s and the range is 0 to 1. There's no conversion; the values are applied as-is. (Aside from Ns, where the range is 0 to 1000, so it's divided by 1000 since _Shininess is 0 to 1.)

--Eric

 

Ok thanks... I wonder what the conversion to Unity's standard shaders is then to make it look more shiny and vibrant. The 1:1 conversion really washs out all the colors of the materials...

 

It might just be your lighting; in my tests the results were as expected.

--Eric

 

Or my .mtl files. I got them from a different vendor and it seems like they exported it differently and now I have to readjust all specular Colors and the shininess because all materials look washed out.

 

Hi Eric -

First of all, wanted to thank you for some of the great tools you've created. Appreciate your work + the support you give your users.

Anyway, I'm running into some CPU spikes - normally these wouldn't be issues but in VR it causes it to kick into "non-vomit-inducing mode" and fades in and out of a different scene.

My question is: is there a way I might be able to throttle CPU usage for remote obj loads - or at last spread out the workload to stay under a certain timing threshold. I'm already running the code in a co-routine and using ConvertFileAsync. The obj files are 250k - 1MB. See attached for profiler stats.

Thoughts appreciated! Thanks.

 

Hi, the async part of ConvertFileAsync refers to downloading a file; the conversion itself always takes place in one frame. At this time there isn't a way to run that over multiple frames, sorry.

--Eric

 

Hello Eric,
I bought the dll and was successful to load and display the obj from my android phone location. But I am having this problem to rotate the model. I tried to use the objRotation feature in the update function but seems like it only rotates once and does not do anything. Can you show me how to constantly have the rotation in the update function. I need this for the obj to rotate according to the head movement in VR. I am very new to unity development as well. Thanks in advance.

 

As the docs say, "ObjRotation will rotate the resulting object meshes around the specified x, y, and z axes. Like ScaleFactor, this 'bakes' the rotation into the mesh, rather than rotating the GameObjects." It can't be used to continuously rotate an object.

The ConvertFile function returns an array of GameObjects (if there's only one object, the array will have one entry). So you can simply refer to the transform component of the GameObject(s) in the array. The Example2-WWW script shows an example of this. If you're new to Unity, I would very strongly recommend spending time doing some tutorials first to become familiar with the engine and API.

--Eric

 

Thanks,
Can you tell me how to get the GameObjects from the ConvertFile. I tried it just like the convertFileAsync

ObjReader.ObjData objData;

objData = ObjReader.use.ConvertFileAsync (objFileName, true, standardMaterial);


and it shows error "cannot implicitly convert unityEngine.Gameobject[] to ObjReader.ObjData

when I try this

objData = ObjReader.use.ConvertFile (objFileName, true, standardMaterial, transparentMaterial);

Thanks a lot again for helping

 

´maybe because you forgot to place the mtl file in the same directory as your obj file. i have the same Problem, works in Editor and not in Android. now it worked.


Btw. Thanks Eric for the Super Assets.
It is my first Projekt with Unity.
I hope that it can load more than 65000 Recticle in the future

 

I got it. I declared a gameobject like

GameObject[] go;


Then in start did

go = ObjReader.use.ConvertFile(objFileName, true, standardMaterial,transparentMaterial);


and in update function I did the following to rotate

go[0].transform.Rotate(Vector3.right * Time.deltaTime);


Thanks. My goal is to use the head movement in VR and rotate the obj according to that. Will work on that now.

 

Hi. I am facing the same issue with a large model (> 100 000) vertices, and I am trying to use groups in the obj file to split up the meshes, but I still get the error " the number of vertices exceeds the maximum limit of 65534 ..."
My question is if it is possible to "restart" the indexing of vertices within the file for each group, or are every new vertex always given a the next index number? I.e. so that in the following example the face indices for group-2 refer to the last "set" of vertices and not the first?

g group-1
v 0 0 0
v 0 10 0
v 0 10 10
f 1 2 3

g group-2
v 20 0 0
v 20 10 0
v 20 10 10
f 1 2 3

Or do I have to split my big obj file into separate files?

 

I realized that it was the "Combine Multiple Groups" flag on the ObjReader that generated the error, as all my groups were merged into one mesh resulting in too many vertices. Setting this flag to false resolved my problem

So it is not a problem that the vertex index is higher than 65,534 as long as the resulting mesh does not contain more that 65,534 vertices, and I guess "restarting" the vertex index in the obj file is not possible and not needed!

 

The vertex index is in fact restarted for each group as long as mesh combining is not used. Hence there's no limit on the number of vertices in a file, as long as each group < 65K vertices.

--Eric

 

HI is there a built in function to centre the pivot point to the centre of the mesh's bounds? Thanks

 

Yes, there's an "auto center on origin" option.

--Eric

 

Hi Eric,

I am using Unity to load an obj file and view it through Oculus in virtual reality.
My problem is that I have around 50000 such obj files which need to be loaded one at a time in the scene.
So basically I want to replace a previous obj file loaded in the scene with a new file during runtime.
The files are named in an order; ex: file_000.obj, file_001.obj, file_002.obj,...etc.

Will the OBJReader solve my problem? If so, how and what functions can I write in my C# code to load these files from my local disk?

 

Sure, that's not a problem. The ConvertFile function in ObjReader reads files from disk.

--Eric

 

Hi Eric, I'm trying to use the ObjReadaer for the first time but I'm getting a null reference exception on the line:

objData = ObjReader.use.ConvertFileAsync (objFileName, true, standardMaterial);

I am sure that objFileName and standardMaterial are NOT null so that just leaves me with the ObjReader. But isn't that all wrapped up in a dll? My original decalration of the objReader looks like this (C#):

public ObjReader.ObjData objData;

Does that all sound okay? Demo all works fine so it must be something in my code.

 

You can try using the source to see if there's anything going on in the ObjReader code. What you posted seems OK from what I can tell.

--Eric

 

Hi!

I have a small feature request. Could you add an option to automatically split a mesh into multiple objects if it exceeds the 65k vertex limit? There is an option to combine OBJ groups into one object, but it would be great if there would be an option to split a mesh too. I know that 65k vertices is a lot, but there are cases, when you have to deal with such a mesh.

Thanks a lot,
Maciej

 

Hi,
I'm looking for an obj-loader that supports objects with more than 65000 vertices, too. Unfortunately, I can't simplify the object or reduce the number of vertices.
regards
CC

 

ObjReader supports objects with >65K vertices if the object is organized into groups where each group <65K vertices.

--Eric

 

Yes, I am aware of that. I was thinking of an object with a single group that exceeds the 65k limit and some kind of auto-splitting mechanism (such as splitting the triangles of a mesh into multiple game objects).

 

Hi Eric,

I imported the ObjReader from Asset Store after which I had no errors but when I clicked extras-> ObjReaderSource.UnityPackage it imported some .cs file and after which I am seeing these two errors:
1> Assets/ObjReader/Demo Scripts/Example1_TextAsset.cs(20,25): error CS0433: The imported type `ObjReader' is defined multiple times
2> Assets/ObjReader/Demo Scripts/Example2-WWW.js(7,23): BCE0018: The name 'ObjReader.ObjData' does not denote a valid type ('not found'). Did you mean 'System.Reflection.Emit.OpCodeType'?

Could you please help me how to remove these errors?

Regards,
Akanksha

 

You can't have both the source and the DLL imported at the same time.

--Eric

 

What shall I do now to remove one of it and make things working?

 

Delete the ObjReader.cs script.

--Eric

 

Hi Eric,

Thanks for the great asset. Everything worked fine until I tried to save my app as WebGL, I'm going through the process of changing

ObjReader.use.ConvertFile

to

ObjReader.use.ConvertFileAsync

and am running into a problem, which I think is more to do with my understanding of C# yield statements.

The following is my import code, it never gets to adding components, am I doing something obviously wrong?

Thanks

Code (CSharp):

1. IEnumerator ImportNow(string path){
2.         Material randomMaterial = materialArray[Random.Range(0,materialArray.Length)];
3.        // GameObject[] importedModels = ObjReader.use.ConvertFile (path, false, randomMaterial); // this works with the non asyncronous
4.  
5.         objData = ObjReader.use.ConvertFileAsync (path, false, randomMaterial);
6.         while (!objData.isDone) {
7.             Debug.Log("loading");
8.             yield return null;
9.          
10.         }
11.         GameObject[] importedModels = objData.gameObjects;
12.  
13.         Debug.Log(path);
14.  
15.         GameObject importedModel = importedModels[0];
16.  
17.         GameObject walls = GameObject.Find("Walls");
18.         importedModel.transform.parent = walls.transform;
19.      
20.         importedModel.AddComponent<BoxCollider>();
21.         importedModel.AddComponent<classInitialConditions2>();
22.         importedModel.AddComponent<classScale2>();
23.         importedModel.AddComponent<classTranslate2>();
24.         importedModel.AddComponent<classRotate2>();
25.         importedModel.AddComponent<classSafeSpace>();
26.         importedModel.AddComponent<AudioSource>();
27.  
28.  
29.  
30.  
31.         do{
32.             rndPos = RandomPositionWithin(limits.transform);
33.             importedModel.transform.localPosition = rndPos;
34.         } while (!IsInsideLimits(importedModel.transform, limits.transform));
35.  
36.      
37.      }

EDIT: A little more background- ImportNow gets called a number of times from a for loop. For some reason the first model thats imported is getting made the child of "Walls" then getting lots of copies of the components added to it, rather than copies added to each imported model ( all models get imported ). Any ideas?