Mesh Match brings AI vision to Maya.
Currently, this core capability powers two main features: landmark detection and skeleton building.
These features operate independently of Mesh Match’s registration functions, meaning you can access them even without a Mesh Match node in your scene. When used standalone, the user interface will adapt to display a single viewport.
AI features are instant regardless of the mesh(es) density.
Automatic Landmarks Detection
Accurately detect facial or body landmarks with customizable settings, and use them as a base for registration, or any application you can think of in the character workflow, or for external use in other softwares.
Automatic Skeleton Build
Create full-body skeletons with a single click, providing a solid foundation for your auto-rigging pipeline.
Options
- AI Model Type:
Face detection can identify up to 468 points. For this tool, there are four presets available: 12, 25, 50 points for faces, or ‘Face + Body’. Using the 12-point preset for faces is recommended. You can then manually add extra points for areas like the ears, skull, and neck if needed. - Perform On:
This allows you to customize on which mesh the AI capture is performed. For the pin pairs addition, you will need to set the menu to ‘Reference And Target’. - Capture Angle:
For more accuracy, the process will take multiple snapshots of the mesh in order to find the regions of interest and keypoints.
The capture angle sets the spread of the capture iterations, around the Y axis of the camera. For example, a value of 50 will make the camera rotate from -25 to +25 degrees around the model. - Capture Steps:
Represents the amount of snapshots that will be performed within the capture angle.
Note: If the results are not accurate (ie skeleton bones are not within the mesh), try to reduce the capture angle.
Custom Commands
You can use the AI vision features independently from the MeshMatch interface by using the meshMatchDetectLandmarks command documented here.
You can also explore the existing implementation by looking at the related functions from the mesh_match_pins_widgets.py and mesh_match_commands.py files in order to implement the AI vision with custom panels and cameras.
Examples of skeletons built with the AI feature