I’m currently trying to figure how to use the ML Deformer between Maya and Unreal. After a couple of tests, I have arrived to promising results for what I’m looking for, except on the translate part.
I did a custom mesh in Maya for testing and added a deformer that make a flare on the base of the mesh when a translation on X occur and a small wave on the base when I rotate on X.
I then use the ML Deformer that I have installed in Maya. Under the tab “UE MLDeformer->Data Generator” to generate the data. I add the translates and the rotations of controller that I need and click the “Generate button”. Once it’s generated, under the tab “UE MLDeformer->Export to Unreal”, I create the FBX and ABC for Unreal.
In Unreal, I import both of the files created previously and I create a brand new “ML Deformer” uasset. I then add the Skeletal Mesh, the anim sequence and the GeoCache in the designated space in the detail panel. I link the bones network inputs with the “Add all animated bones” option. Then I start the training with the model “Neural Morph Model”.
Once the training is finished, I swap from “Training” to “Testing” and set the “Test anim Sequence”, the “Deformer Graph” (DG_Function_LinearBlendSkin_Morph), the “Ground Truth” and activate the heat map checkbox. When I scrub the timeline everything related to the rotations seems ok, even if it is not 100% accurate with the ground truth, but everything related to the translate does not apply. Ironically, I can see that it is supposed to change when I go back to “Training” since I can see the green dots that sculpt the deformation.
Does anyone know why nothing happens? I’m pretty sure it is a param issue since I can see the dots but even if I change the values multiple times, the results is always the same.