Hi, I am really confused about the transAffine and transAffineNorm regularizers. I read the original SSDR paper and some following papers like the Robust SSDR in 2014, but can't find the reference for transAffineNorm regularizers.

Would you please tell the reference about the transAffine and transAffineNorm regularizers? Thanks a lot!

Occasionally I run in to cases where DemBones gets stuck in an infinite loop while trying to initialize the bones in DemBones.h init(). I've attached a sample fbx and abc that causes the issue.

faceshapes.zip

Any insight as to why?

../bin/DemBones.exe -i="neutral.fbx" -a="face_rom.abc" -b=64 -o="Decomposition_64.fbx"
1 animated mesh sequence(s):
    "face_rom.abc" + "neutral.fbx" --> "Decomposition_64.fbx"
Parameters:
    nBones (target)    = 64
    nInitIters         = 10 (default)
    nIters             = 30 (default)
    nTransIters        = 5 (default)
    nWeightsIters      = 3 (default)
    bindUpdate         = 0 (no update) (default)
    transAffine        = 10 (default)
    transAffineNorm    = 4 (default)
    nnz                = 8 (default)
    weightsSmooth      = 0.0001 (default)
    weightsSmoothStep  = 1 (default)
Checking ABCs:
    "face_rom.abc": /Louise_asym/Louise_asymShape (57 frames)
Reading ABCs......................................................... Done!
Reading FBXs:
    "neutral.fbx"... Done!
    5034 vertices
Initializing bones: 1>1>1>1>1>1>1>1>1>1>1>1>1>1>1>1>1>1>1>1>1>1>1>1>1>1>1>1>1>1>1>1>1>1>1>1>1>1>1>1>1>1>1>1>1>1>1>1>1>1>1>1>1>1>1>1>1>1>1>1>1>1>1>1>1>1>1>1>1>1>1>1>1>1>1>1>1>1>1>

Is it possible for Dem-Bones to directly convert a sequence of PLY or OBJ files into a single animated and skinned FBX, instead of using an animated Alembic files?

Hi, Thanks for your code! Really nice work! I have one question about Best Rigid Transformation from covariance matrix. From the paper here https://igl.ethz.ch/projects/ARAP/svd_rot.pdf or the SSDR paper

in the function qpT2m:

Eigen::JacobiSVD svd(...) d(2, 2) = (svd.matrixU()*svd.matrixV.transpose()).determinant(); m.rotMat(k, j) = svd.matrixU()dsvd.matrixV().transpose();

should be ?

d(2, 2) = (svd.matrixV()*svd.matrixU.transpose()).determinant(); m.rotMat(k, j) = svd.matrixV()dsvd.matrixU().transpose();

=================================== but definitly from the result, your implementation is right. I'm really confused!

Thanks!

I compile in debug mode in VS2019, Win10. all configures are in debug mode when configure cmake-gui. FBXSDK is debug version lib.

but it seems Alembic from Alembic (from Maya 2020 Update 2 DevKit) only contains release libs. How to get Debug version Alembic? Should I compile Alembic from https://github.com/alembic/alembic? Then how to organize the catalogue? Which version of Alembic (from Maya 2020 Update 2 DevKit) is?

The compile error is: 1>Alembic.lib(DataTypeRegistry.obj) : error LNK2038: mismatch detected for 'RuntimeLibrary': value 'MD_DynamicRelease' doesn't match value 'MDd_DynamicDebug' in AbcReader.obj 1>Alembic.lib(SprImpl.cpp.obj) : error LNK2038: mismatch detected for '_ITERATOR_DEBUG_LEVEL': value '0' doesn't match value '2' in AbcReader.obj 1>Alembic.lib(SprImpl.cpp.obj) : error LNK2038: mismatch detected for 'RuntimeLibrary': value 'MD_DynamicRelease' doesn't match value 'MDd_DynamicDebug' in AbcReader.obj 1>Alembic.lib(AprImpl.cpp.obj) : error LNK2038: mismatch detected for '_ITERATOR_DEBUG_LEVEL': value '0' doesn't match value '2' in AbcReader.obj 1>Alembic.lib(AprImpl.cpp.obj) : error LNK2038: mismatch detected for 'RuntimeLibrary': value 'MD_DynamicRelease' doesn't match value 'MDd_DynamicDebug' in AbcReader.obj 1>Alembic.lib(ScalarPropertyWriter.obj) : error LNK2038: mismatch detected for '_ITERATOR_DEBUG_LEVEL': value '0' doesn't match value '2' in AbcReader.obj 1>Alembic.lib(ScalarPropertyWriter.obj) : error LNK2038: mismatch detected for 'RuntimeLibrary': value 'MD_DynamicRelease' doesn't match value 'MDd_DynamicDebug' in AbcReader.obj 1>Alembic.lib(ArrayPropertyWriter.obj) : error LNK2038: mismatch detected for '_ITERATOR_DEBUG_LEVEL': value '0' doesn't match value '2' in AbcReader.obj 1>Alembic.lib(ArrayPropertyWriter.obj) : error LNK2038: mismatch detected for 'RuntimeLibrary': value 'MD_DynamicRelease' doesn't match value 'MDd_DynamicDebug' in AbcReader.obj 1>Alembic.lib(ArrayPropertyReader.obj) : error LNK2038: mismatch detected for '_ITERATOR_DEBUG_LEVEL': value '0' doesn't match value '2' in AbcReader.obj 1>Alembic.lib(ArrayPropertyReader.obj) : error LNK2038: mismatch detected for 'RuntimeLibrary': value 'MD_DynamicRelease' doesn't match value 'MDd_DynamicDebug' in AbcReader.obj 1>Alembic.lib(ScalarSample.obj) : error LNK2038: mismatch detected for '_ITERATOR_DEBUG_LEVEL': value '0' doesn't match value '2' in AbcReader.obj 1>Alembic.lib(ScalarSample.obj) : error LNK2038: mismatch detected for 'RuntimeLibrary': value 'MD_DynamicRelease' doesn't match value 'MDd_DynamicDebug' in AbcReader.obj 1>AlembicHalf.lib(half.obj) : error LNK2038: mismatch detected for '_ITERATOR_DEBUG_LEVEL': value '0' doesn't match value '2' in AbcReader.obj 1>AlembicHalf.lib(half.obj) : error LNK2038: mismatch detected for 'RuntimeLibrary': value 'MD_DynamicRelease' doesn't match value 'MDd_DynamicDebug' in AbcReader.obj 1>AlembicIex.lib(IexBaseExc.obj) : error LNK2038: mismatch detected for '_ITERATOR_DEBUG_LEVEL': value '0' doesn't match value '2' in AbcReader.obj 1>AlembicIex.lib(IexBaseExc.obj) : error LNK2038: mismatch detected for 'RuntimeLibrary': value 'MD_DynamicRelease' doesn't match value 'MDd_DynamicDebug' in AbcReader.obj 1>LINK : warning LNK4098: defaultlib 'MSVCRT' conflicts with use of other libs; use /NODEFAULTLIB:library 1>D:\Projects\dem-bones\build_debug\Debug\DemBones.exe : fatal error LNK1319: 184 mismatches detected 1>Done building project "DemBonesCmd.vcxproj" -- FAILED. ========== Build: 0 succeeded, 1 failed, 1 up-to-date, 0 skipped ==========

There are finished skin meshes and animations, and then I added some blend shapes. I added some new bones at the origin without animation, and set lockM to 0 for those new bones. (The original bone is set to 1) Set the lockW value of the deformation vertex to 0. (set Others to 1) Set matrix w and matrix m with original data，set w value of new bones to 0, and m value of new bones to Matrix4::Identity(). I want these bones to replace blend shapes，but it did not work. The result of the compute is that these bones are still at the origin, and there are no vertices with their weights. I don't know how to work properly

Hi and thanks a lot for this, it works very well !

I was wondering if it could be possible to allow joint scale to better match the deformed geometry. I think, in some cases where scale is supported, it could help reduce the number of bone needed.

Thanks again!

Hi, I recently used SSDR created a skeleton animation and write the animation into an .fbx file. Then I tried to use fbxsdk to extract the translation and rotation data of each bone from the animation.

However, I found that the extracted translation or rotation(both Global or Local ) data are always the same as the initial value, but when I import the fbx file into Maya, the animation looks fine, and the local translation and rotation value of each bone varies over time. The data extraction code are shown below:

One thing wierd is that I found the number of AnimStack in the .fbx which created by SSDR are always 2. I'm not sure whether this factor leads to my problem.

bool readFbx(string file_path)
{
	FbxManager* Sdkmanager = NULL;
	fbxsdk::FbxScene* Scene = NULL;
	bool Result;
	//prepare the FBX SDK
	InitializeSdkObjects(Sdkmanager, Scene);
	FbxString file(file_path.c_str());
	Result = LoadScene(Sdkmanager, Scene, file.Buffer());
	vector<FbxString>boneName;
	vector<fbxsdk::FbxNode*>bone_Node;
	if (Result == false)
	{
		FBXSDK_printf("\n\nAn error occurred while loading the scene...");
		return Result;
	}
        else
	{
		// get the name of each bones of the skeleton
		fbxsdk::FbxNode* root = Scene->GetRootNode();
		for (int i = 0; i < root->GetChildCount(); ++i)
		{
			fbxsdk::FbxNode* pNode = root->GetChild(i);
			fbxsdk::FbxNodeAttribute::EType lAttributeType;
			if (pNode->GetNodeAttribute() == NULL)
			{
				FBXSDK_printf("NULL Node Attribute\n\n");
			}
			else
			{
				lAttributeType = (pNode->GetNodeAttribute()->GetAttributeType());
				switch (lAttributeType)
				{
				case fbxsdk::FbxNodeAttribute::eSkeleton:
					GetSkeletonInformation(boneName, bone_Node, pNode);//遍历节点找到mesh节点
					break;
				case FbxNodeAttribute::eCamera:
					break;
				default:
					break;
				}
			}
		}
                // read the global translation data over time
		const int AnimStackCount = Scene->GetSrcObjectCount<fbxsdk::FbxAnimStack>();
		for (int i = 0; i < AnimStackCount; ++i)
		{
			fbxsdk::FbxAnimStack* pstack = Scene->GetSrcObject<fbxsdk::FbxAnimStack>(i);
			const char* stackname = pstack->GetName();
			cout << "Stack: " << stackname << endl;

			const int nLayerCount = pstack->GetMemberCount();
			for (int j = 0; j < nLayerCount; ++j)
			{
				fbxsdk::FbxAnimLayer* pLayer = pstack->GetMember<fbxsdk::FbxAnimLayer>(j);
				const char* layername = pLayer->GetName();
				cout << "Layer: " << layername << endl;
				fbxsdk::FbxAnimCurve* lCurveRX = bone_Node[0]->LclRotation.GetCurve(pLayer, FBXSDK_CURVENODE_COMPONENT_X, false);
				if (lCurveRX == NULL)
				{
					continue;
				}
				int pRXCount = lCurveRX->KeyGetCount();
				for (int s = 0; s < pRXCount; ++s)
				{
					FbxTime ktime = lCurveRX->KeyGetTime(s);
					vector<float> bone_position;
					for (int bone = 0; bone < bone_Node.size(); ++bone)
					{
						FbxAMatrix GLOBAL = bone_Node[bone]->EvaluateLocalTransform(ktime);
						FbxVector4 T = GLOBAL.GetT();
						
						fbxsdk::FbxString tmpName("LiYuanQian_rig_v001:Root_M");
						if (boneName[bone] == tmpName)
						{
							cout << tmpName.Buffer() << "frame: " << s << " trans:" << T[0] << "," << T[1] << "," << T[2] << "." << " ktime"  << endl;;
						}
						
						for (int k = 0; k < 3; ++k)
						{
							bone_position.push_back(T[k]);
							
						}
					}
				}

			}
		}
		return Result;
	}
}

Thanks sincerely for sharing your code on Github, But I still have a few confusion. I don't see any code implementing bone joints constraints. I don't know if I missed something or you have not updated it yet.

Hello,

I am getting the following error messages when I compile demBones in VS2019:

Error: Dem::DemBones: use of class templates requires template argument list

Error C2352 'Dem::DemBones<_Scalar,_AniMeshScalar>::clear': illegal call of non-static member function cmt C:\Users\max\source\repos\cmt\src\DemBones\DemBonesExt.h 108

The line error is in bold text.

DemBonesExt.h ... /** @brief Clear all data */ void clear() { fTime.resize(0); boneName.resize(0); parent.resize(0); bind.resize(0, 0); preMulInv.resize(0, 0); rotOrder.resize(0, 0); DemBones::clear(); }

Any idea? Thx, Max

gcc 7.4.0 reports an warning to this line

https://github.com/electronicarts/dem-bones/blob/2ab9b0447ad4b8bf4a141b9c65af717e7e2c2e0f/include/DemBones/DemBones.h#L297

Returning reference to temporary triggers segmentation fault or undefined behavior. Work around is creating an variable.

        VectorX x0 = w.col(i).toDense().cwiseMax(0.0);
        x=indexing_vector(x0, idx.head(nnzi));

Although it is better to fix not returning temp address in Indexing.h

/home/syoyo/work/dem-bones/include/DemBones/Indexing.h: In instantiation of ‘const typename ArgType::Scalar& Dem::indexing_functor_vector<ArgType, IndexType>::operator()(Eigen::Index) const [with ArgType = Eigen::CwiseBinaryOp<Eigen::internal::scalar_max_op<double, double>, const Eigen::Matrix<double, -1, 1>, const Eigen::CwiseNullaryOp<Eigen::internal::scalar_constant_op<double>, Eigen::Matrix<double, -1, 1> > >; IndexType = Eigen::VectorBlock<Eigen::Array<int, -1, 1>, -1>; typename ArgType::Scalar = double; Eigen::Index = long int]’:
/home/syoyo/work/dem-bones/ExtLibs/Eigen/Eigen/src/Core/CoreEvaluators.h:430:110:   required from ‘Scalar Eigen::internal::nullary_wrapper<Scalar, NullaryOp, false, true, false>::operator()(const NullaryOp&, IndexType) const [with IndexType = long int; Scalar = double; NullaryOp = Dem::indexing_functor_vector<Eigen::CwiseBinaryOp<Eigen::internal::scalar_max_op<double, double>, const Eigen::Matrix<double, -1, 1>, const Eigen::CwiseNullaryOp<Eigen::internal::scalar_constant_op<double>, Eigen::Matrix<double, -1, 1> > >, Eigen::VectorBlock<Eigen::Array<int, -1, 1>, -1> >]’
/home/syoyo/work/dem-bones/ExtLibs/Eigen/Eigen/src/Core/CoreEvaluators.h:533:37:   required from ‘Eigen::internal::evaluator<Eigen::CwiseNullaryOp<NullaryOp, PlainObjectType> >::CoeffReturnType Eigen::internal::evaluator<Eigen::CwiseNullaryOp<NullaryOp, PlainObjectType> >::coeff(IndexType) const [with IndexType = long int; NullaryOp = Dem::indexing_functor_vector<Eigen::CwiseBinaryOp<Eigen::internal::scalar_max_op<double, double>, const Eigen::Matrix<double, -1, 1>, const Eigen::CwiseNullaryOp<Eigen::internal::scalar_constant_op<double>, Eigen::Matrix<double, -1, 1> > >, Eigen::VectorBlock<Eigen::Array<int, -1, 1>, -1> >; PlainObjectType = Eigen::Matrix<double, -1, 1>; Eigen::internal::evaluator<Eigen::CwiseNullaryOp<NullaryOp, PlainObjectType> >::CoeffReturnType = double]’
/home/syoyo/work/dem-bones/ExtLibs/Eigen/Eigen/src/Core/AssignEvaluator.h:642:5:   required from ‘void Eigen::internal::generic_dense_assignment_kernel<DstEvaluatorTypeT, SrcEvaluatorTypeT, Functor, Version>::assignCoeff(Eigen::Index) [with DstEvaluatorTypeT = Eigen::internal::evaluator<Eigen::Matrix<double, -1, 1> >; SrcEvaluatorTypeT = Eigen::internal::evaluator<Eigen::CwiseNullaryOp<Dem::indexing_functor_vector<Eigen::CwiseBinaryOp<Eigen::internal::scalar_max_op<double, double>, const Eigen::Matrix<double, -1, 1>, const Eigen::CwiseNullaryOp<Eigen::internal::scalar_constant_op<double>, Eigen::Matrix<double, -1, 1> > >, Eigen::VectorBlock<Eigen::Array<int, -1, 1>, -1> >, Eigen::Matrix<double, -1, 1> > >; Functor = Eigen::internal::assign_op<double, double>; int Version = 0; Eigen::Index = long int]’
/home/syoyo/work/dem-bones/ExtLibs/Eigen/Eigen/src/Core/AssignEvaluator.h:501:7:   required from ‘static void Eigen::internal::dense_assignment_loop<Kernel, 1, 0>::run(Kernel&) [with Kernel = Eigen::internal::generic_dense_assignment_kernel<Eigen::internal::evaluator<Eigen::Matrix<double, -1, 1> >, Eigen::internal::evaluator<Eigen::CwiseNullaryOp<Dem::indexing_functor_vector<Eigen::CwiseBinaryOp<Eigen::internal::scalar_max_op<double, double>, const Eigen::Matrix<double, -1, 1>, const Eigen::CwiseNullaryOp<Eigen::internal::scalar_constant_op<double>, Eigen::Matrix<double, -1, 1> > >, Eigen::VectorBlock<Eigen::Array<int, -1, 1>, -1> >, Eigen::Matrix<double, -1, 1> > >, Eigen::internal::assign_op<double, double>, 0>]’
/home/syoyo/work/dem-bones/ExtLibs/Eigen/Eigen/src/Core/AssignEvaluator.h:767:37:   required from ‘void Eigen::internal::call_dense_assignment_loop(DstXprType&, const SrcXprType&, const Functor&) [with DstXprType = Eigen::Matrix<double, -1, 1>; SrcXprType = Eigen::CwiseNullaryOp<Dem::indexing_functor_vector<Eigen::CwiseBinaryOp<Eigen::internal::scalar_max_op<double, double>, const Eigen::Matrix<double, -1, 1>, const Eigen::CwiseNullaryOp<Eigen::internal::scalar_constant_op<double>, Eigen::Matrix<double, -1, 1> > >, Eigen::VectorBlock<Eigen::Array<int, -1, 1>, -1> >, Eigen::Matrix<double, -1, 1> >; Functor = Eigen::internal::assign_op<double, double>]’
/home/syoyo/work/dem-bones/ExtLibs/Eigen/Eigen/src/Core/AssignEvaluator.h:926:31:   [ skipping 3 instantiation contexts, use -ftemplate-backtrace-limit=0 to disable ]
/home/syoyo/work/dem-bones/ExtLibs/Eigen/Eigen/src/Core/AssignEvaluator.h:808:18:   required from ‘void Eigen::internal::call_assignment(Dst&, const Src&) [with Dst = Eigen::Matrix<double, -1, 1>; Src = Eigen::CwiseNullaryOp<Dem::indexing_functor_vector<Eigen::CwiseBinaryOp<Eigen::internal::scalar_max_op<double, double>, const Eigen::Matrix<double, -1, 1>, const Eigen::CwiseNullaryOp<Eigen::internal::scalar_constant_op<double>, Eigen::Matrix<double, -1, 1> > >, Eigen::VectorBlock<Eigen::Array<int, -1, 1>, -1> >, Eigen::Matrix<double, -1, 1> >]’
/home/syoyo/work/dem-bones/ExtLibs/Eigen/Eigen/src/Core/PlainObjectBase.h:779:32:   required from ‘Derived& Eigen::PlainObjectBase<Derived>::_set(const Eigen::DenseBase<OtherDerived>&) [with OtherDerived = Eigen::CwiseNullaryOp<Dem::indexing_functor_vector<Eigen::CwiseBinaryOp<Eigen::internal::scalar_max_op<double, double>, const Eigen::Matrix<double, -1, 1>, const Eigen::CwiseNullaryOp<Eigen::internal::scalar_constant_op<double>, Eigen::Matrix<double, -1, 1> > >, Eigen::VectorBlock<Eigen::Array<int, -1, 1>, -1> >, Eigen::Matrix<double, -1, 1> >; Derived = Eigen::Matrix<double, -1, 1>]’
/home/syoyo/work/dem-bones/ExtLibs/Eigen/Eigen/src/Core/Matrix.h:225:24:   required from ‘Eigen::Matrix<_Scalar, _Rows, _Cols, _Options, _MaxRows, _MaxCols>& Eigen::Matrix<_Scalar, _Rows, _Cols, _Options, _MaxRows, _MaxCols>::operator=(const Eigen::DenseBase<OtherDerived>&) [with OtherDerived = Eigen::CwiseNullaryOp<Dem::indexing_functor_vector<Eigen::CwiseBinaryOp<Eigen::internal::scalar_max_op<double, double>, const Eigen::Matrix<double, -1, 1>, const Eigen::CwiseNullaryOp<Eigen::internal::scalar_constant_op<double>, Eigen::Matrix<double, -1, 1> > >, Eigen::VectorBlock<Eigen::Array<int, -1, 1>, -1> >, Eigen::Matrix<double, -1, 1> >; _Scalar = double; int _Rows = -1; int _Cols = 1; int _Options = 0; int _MaxRows = -1; int _MaxCols = 1]’
/home/syoyo/work/dem-bones/include/DemBones/DemBones.h:303:6:   required from ‘void Dem::DemBones<_Scalar, _AniMeshScalar>::computeWeights() [with _Scalar = double; _AniMeshScalar = float]’
/home/syoyo/work/dem-bones/include/DemBones/DemBones.h:344:18:   required from ‘void Dem::DemBones<_Scalar, _AniMeshScalar>::compute() [with _Scalar = double; _AniMeshScalar = float]’
/home/syoyo/work/dem-bones/src/command/mainCmd.cpp:171:16:   required from here
/home/syoyo/work/dem-bones/include/DemBones/Indexing.h:106:30: warning: returning reference to temporary [-Wreturn-local-addr]
   return m_arg(m_indices[idx]);

Initializing bones: 1>1 Only 1 bone was generated in .fbx file Is there anything not set? How to generate 20 bones normally。 Looking forward to your reply

In class MyDemBones of mainCmd.cpp, the member variables tolerance and patience are initialized to (1e-3) and (3).

However, these values are never updated to those passed in via the command line.

It looks like the following lines must be added to the main function to ensure that the user-provided arguments are being respected:

model.tolerance = dbOptions.tolerance;
model.patience = dbOptions.patience;

I have some .abc files and the init pose .fbx.

Do I need to concatenate all the motion sequences into one .abc file? or I can use some scripts to decompose multi sequences in the meantime?

The script I used: call "../bin/Windows/DemBones.exe" -i="dress.fbx" -a="dress_shape00_01_01.abc" -b=40 -o="dress_40.fbx" -n=100 --nWeightsIters=3 -z=8

The decomposition result looks terrible. Can you give some suggestions?

Hi, First, I want to say that I'm very impressed with DemBones solution and its possibilities. I'm wondering if it is possible to obtain the output animation of the RIG with changed only the rotation in the joints (translation would be only applied to the root). From the documentation of the parameters, it is not obvious to me how to achieve this. There is a demLockparameter, but it seems to lock the whole transformation of the joint in the animation.

The issue is as :

mac@Ayushs-MacBook-Pro MacOS % ./dembones -a=a.abc -i=a.fbx -o=out.fbx -b=16 dyld: Library not loaded: /usr/local/opt/llvm/lib/libomp.dylib Referenced from: /Users/mac/Desktop/dem-bones-1.2.0/bin/MacOS/./dembones Reason: no suitable image found. Did find: /usr/local/opt/llvm/lib/libomp.dylib: mach-o, but wrong architecture /usr/local/opt/llvm/lib/libomp.dylib: mach-o, but wrong architecture zsh: abort ./dembones -a=a.abc -i=a.fbx -o=out.fbx -b=16