Create an ML Deformer using separate target geometry

Important: When you train the ML Deformer, you may be prompted to download and install external Python modules. By default, the install path of the extra modules is <MAYA_APP_DIR>/mlDeformer/<ml_deformer_version>/Lib/site-packages. If you need to download these modules manually and/or install them to a custom location, see Install Python libraries to train the ML Deformer.

This topic gives a second example of how the ML Deformer simulates deformation animation from one deformed object on another object. You can also create an ML Deformer using the Deformer Attributes widget (Deformer stack), which is useful if you have a rig already set up as it avoids the work transferring the deformation target mesh. See Create an ML Deformer.

To illustrate the process, this scene has two meshes, a static mesh (Disc 1) and one with a Cluster deformer (Disc 2). We will train the ML Deformer with the animation from Disc 2 so we can use it on Disc 1. Disc 2 is the Target deformation, meaning it is the goal of what the ML Deformer is simulating.

Disc 1 is a static mesh (Disc 1) and (Disc 2) has a Cluster deformer.

Tip: You can test and experiment with ML Deformer sample files in the Content Browser (Windows > Content Browser > Examples > Animation > MLDeformer).

Sample ML Deformer animation in the Content Browser

To add an ML Deformer to geometry

    Important: Before you start: Training the ML model requires that you download external Python modules. See step #11 for more information.
  1. Select the object that you want to receive the deformation. In this example, Disc 1.
  2. Select Deform > ML Deformer in the Animation (F4), Rigging (F3), or Modeling (F2) menu set to add an ML Deformer to Disc 1.
  3. Open the Attribute Editor ML Deformer tab and pin the tab.
  4. Select the geometry with the deformation (Disc 2), and in the Attribute Editor, right-click the Target Geometry column in the ML Deformer table and select Replace Target with Selected.
  5. Switch to the Attribute Editor Control Collector tab and expand the Add Controls table.
  6. Select Disc 2, the Target, and choose the attributes that you want to control the deformation in the Attribute Editor Add Controls list Scalars Tab. (The Matrices tab is used for when you have a deep joint hierarchy. See the Matrices tab section in the ML Deformer Control Collector attributes topic.)

    Ctrl or Shift + click to select multiple or all attributes and click the Add () button at the top right of the list. For this example, as the animation is only on the Y-axis, we select only the translateY attributes.

    Note: You can also right-click and choose either Add Selected or Add All.

    The attributes are added to the Connected Controls list, below the Add Controls list.

  7. Set Min and Max values for the deformation training. In this example, the Min and Max values are set to -50 and 50 because in the example this is a good representation of the animation that we want to simulate.
    Tip: You may need to expand the Attribute Editor window to view the Min and Max columns.
  8. Select the Target geometry (Disc 2). In the Attribute Editor ML Deformer tab, click Export Training Data… to open the Export Training Data window, where you can configure what animation from the Target geometry is used for the ML Deformer. (You can also right-click the ML Model column, and select Export Training Data… from the ML Model contextual menu in the ML Deformer attributes topic.)
  9. In the Export Training Data window, set the following parameters:
    • Training Data Location: Browse to where you want to save your training data and create a folder for it.
    • Export Start/End Frame: Set the range of frames to capture from the Target geometry deformation.
    • Training Data Name: Choose a name for the data.
    • Click Export. The Target object (Disc 2) animation is exported as training data.

      • See ML Deformer Export Training Data window for a detailed explanation of each setting in this window.

  10. In the Attribute Editor ML Deformer tab, click Train the Model… to open the Training ML Deformer window. The Training Data Location field shows where the training data and model are found.

    You can also right-click the ML Model cell in the ML Deformer Attributes list and select Train Model….) See ML Deformer Training Settings for a detailed explanation of each setting in this window.

  11. In the Training ML Deformer window, set the following:
    • Output ML Model Name: Choose a name for your trained model.
    • Activate Preload Data: to speed up the training (at the cost of memory usage).
    • Click Train.
      Two Python windows appear. You can track the progress of this and any other ML Training in the ML Deformer Job Monitor . When you see the message [ML Deformer] Training complete in the Python window, the training is finished.
      Important: When you click Train, you may be prompted to download and install external Python modules. By default, the install path of the extra modules is <MAYA_APP_DIR>/mlDeformer/<ml_deformer_version>/Lib/site-packages. If you need to download these modules manually and/or install them to a custom location, see Install Python libraries to train the ML Deformer.
  12. In the ML Deformer Attribute Editor tab, right-click the ML Model cell in the ML Deformer table and select Select Model > <your model>.
  13. Turn on Enabled in the ML Deformer tab so that the ML Deformer applies the learned deformations.

    Like other deformers, you can adjust the Envelope setting to reduce the effect of the ML Deformation.

Parent topic: ML Deformer