MetaHuman Animator crash - dual camera Stereo HMC

[Newimo]

Hi there,

Can someone please help us get past this crash? We’re very keen to get the Stereo HMC workflow going!

In the process of importing Stereo HMC footage (dual cameras, not iPhone) for MetaHuman Animator, UE 5.2 crashes with the following message:

Assertion failed: Child [File:[D:\build\U5M-Marketplace\Sync\LocalBuilds\PluginTemp\HostProject\Plugins\MetaHuman\Source\MetaHumanPipeline\Public\Pipeline\DataTree.h]](file:///D:/build/U5M-Marketplace/Sync/LocalBuilds/PluginTemp/HostProject/Plugins/MetaHuman/Source/MetaHumanPipeline/Public/Pipeline/DataTree.h]) [Line: 81] Data does not exist (See full error below).

The weird thing is, the above file path does not exist on my workstation, as if it’s been hard-coded into MetaHuman Animator plugin (I don’t have a D: drive).

When I search my computer, ‘DataTree.h’ is not found under UE 5.2 (only in UE 5.1 in a different location).

Just to clarify, I was doing the stereo HMC workflow, following the guide at this link. We completed all of the processing and calibration steps with success.

This is how we reached the crash point:

1. Create blank project
2. Enable MetaHuman Plugin
3. Create a new capture source
4. Selected Stereo HMC Archives
5. Selected a storage path to my ‘data_prepared’ folder
6. Saved
7. Opened the Capture Manager under Tools
8. Selected Capture Source
9. Selected my videos
10. Clicked Add To Queue
11. Clicked Import All
12. CRASH (See crash logs attached)

Assertion failed: Child [File:D:\build\U5M-Marketplace\Sync\LocalBuilds\PluginTemp\HostProject\Plugins\MetaHuman\Source\MetaHumanPipeline\Public\Pipeline\DataTree.h] [Line: 81] Data does not exist

• UnrealEditor_MetaHumanPipeline!UE::MetaHuman::Pipeline::FDataTree::GetData()

UnrealEditor_MetaHumanPipeline!UE::MetaHuman::Pipeline::FPipelineData::GetExitStatus()
UnrealEditor_MetaHumanCaptureSource!FFileFootageIngest::ProcessComplete()
UnrealEditor_MetaHumanCaptureSource!TBaseRawMethodDelegateInstance<0,FFileFootageIngest,void __cdecl(TSharedPtrUE::MetaHuman::Pipeline::FPipelineData,1),FDefaultDelegateUserPolicy,TPromise<TResult<void,FMetaHumanCaptureError> > * __ptr64>::ExecuteIfSafe()
UnrealEditor_MetaHumanPipeline!TMulticastDelegateBase::Broadcast<IBaseDelegateInstance<void __cdecl(TSharedPtrUE::MetaHuman::Pipeline::FPipelineData,1),FDefaultDelegateUserPolicy>,TDelegate<void __cdecl(TSharedPtr<UE::MetaHuman:()
UnrealEditor_MetaHumanPipeline!TMulticastDelegate<void __cdecl(TSharedPtrUE::MetaHuman::Pipeline::FPipelineData,1),FDefaultDelegateUserPolicy>::Broadcast()
UnrealEditor_MetaHumanPipeline!UMetaHumanPipelineProcess::ProcessComplete()
UnrealEditor_MetaHumanPipeline!TBaseUObjectMethodDelegateInstance<0,UMetaHumanPipelineProcess,void __cdecl(TSharedPtrUE::MetaHuman::Pipeline::FPipelineData,1),FDefaultDelegateUserPolicy>::ExecuteIfSafe()
UnrealEditor_MetaHumanPipeline!TMulticastDelegateBase::Broadcast<IBaseDelegateInstance<void __cdecl(TSharedPtrUE::MetaHuman::Pipeline::FPipelineData,1),FDefaultDelegateUserPolicy>,TDelegate<void __cdecl(TSharedPtr<UE::MetaHuman:()
UnrealEditor_MetaHumanPipeline!TMulticastDelegate<void __cdecl(TSharedPtrUE::MetaHuman::Pipeline::FPipelineData,1),FDefaultDelegateUserPolicy>::Broadcast()
UnrealEditor_MetaHumanPipeline!<lambda_be3c7a9f4837aa078e2999ede1c35f6f>::operator()()
UnrealEditor_Core
UnrealEditor_Core
UnrealEditor_Core
UnrealEditor_Engine
UnrealEditor
UnrealEditor
UnrealEditor
UnrealEditor
UnrealEditor
UnrealEditor
kernel32
ntdll

[Kirillspec83]

Could you tell what equipment you use? i mean what type of HMC?

[Newimo]

It seems we had a VRAM memory problem (RTX 3070 8GB) for this particular take.

For this take the solution was to crop the HD video footage from the dual cameras. We went from 1080 x 1920 down to 970 x 1220, no quality loss on the face region.

There’s a possibility that FHD may work with other calibration takes (memory seems to be allocated based on calibration results).

Another option (not tested) is to scale the footage from 1080 to 720, which will elliminate the cropping step from our workflow. However, we’re unsure if the face tracking will suffer quality loss.

We’re using a custom HMC with two drone cameras running FHD @ 60 FPS.

[Kirillspec83]

Greetings!)) I am also using custom HMC with two simple identical action cameras))
i tried 1920x1080 30FPS and it works good on my RTX2060 8GB