{New in version 1.07.00} 


The 'AI Auto Guides Placements' tool help to automatically position the guides on a humanoid character using Machine Learning (ML) pose detection models.


To access the tool, navigate to: 'Edit' menu → 'Guides' → 'AI Auto Guides Placement'


The AI tool operates by analyzing images captured as the camera rotates around the character. It identifies 'landmarks' in each image, and then calculates the most accurate average of those positions in 3D space.


Requirements:

  • Maya 2023 or higher
  • An active Arise license

Upon use, Arise will automatically install any required open-source Python libraries.


The AI tool utilizes three different models for positioning:

  • Body model - Positions nodes such as: (L_/R_) CA_Leg, (L_/R_) CA_arm, CA_Spine, and CA_Head.
  • Fingers model - Positions nodes such as: (L_/R_) CA_Fingers.
  • Face model - Positions nodes such as: (L_/R_) CA_Eye and CA_Mouth.


UI overview:

(Click to enlarge)


Viewport:

The top section contains a viewport linked to a rotating camera focused on the character model. Adjust the viewport to target relevant body parts depending on the selected model:


  • Body Model: Center the entire body.
  • Fingers Model: Center both hands.
  • Face Model: Center the head.


For optimal results, maximize the tool window size to capture larger screenshots, which enhances AI accuracy.


Timeline Slider:

Below the viewport is a slider controlling the timeline, showing you the rotation around the body parts the AI will analyze.

Ensure body parts remain within the frame as the camera rotates.


Frame Count:

Located next to the slider, this count determines the AI’s precision. Higher frame counts improve accuracy but increase computation time.


Model Tabs:

The bottom half of the UI contains tabs for each model: Body, Fingers, and Face.

  • The top button of each tab calculate the landmarks of each model. After centering the character either whole-body, hands, or face in the viewport and setting the frame count, you can press this button to let the AI calculate it's landmarks.
  • To view the calculated landmarks you can click on "Create Display Locators" and a locator will be created for each landmark.
  • The dropdowns for the different body parts will be populated with only allowed nodes and will automatically try and choose the appropriate node.
  • Next to the dropdowns is a 'refresh' button to repopulate the dropdowns with any new nodes added to Arise.
  • On the right side of each tab, model-specific attributes can be adjusted, such as the number of spine or eye landmarks to calculate.
  • The 'Transfer To Body/Fingers/face Guides' button takes the positions of calculated landmarks and transfers them to the guides of the nodes specified in the dropdowns.



Notes:

  • If you see the warning 'No landmarks identified,' it means the AI could not detect a character in the images. Try the tips below to improve the AI's results.
  • The AI analyzes the screenshot of your model to identify character features. Therefore, you should apply textures to your model, hide unnecessary models, smooth the model if needed, or make other adjustments to make the model look complete/realistic.
  • In certain cases, a model's texture can interfere with the AI's ability to detect landmarks. In these situations, switching to a shaded display without textures (by pressing 5 on the keyboard) can improve detection accuracy.
  • Changing the background to create more contrast between the character and its surroundings can help improve AI accuracy. To do this, press 'Alt' + 'B' to adjust the background color.
  • The open-source (Apache License Version 2.0) AI models are trained on consented images of real human images and therefore will provide better results on realistic humanoid characters, but good results have been successfully tested on stylized cartoon humanoids too.
  • The bigger the viewport the larger the image the AI can analyze and therefore provide better results, therefore, make the AI tool window as big as possible.
  • More frames give more samples for the AI to analyze, but takes longer to compute.
  • If you use the 'CA_Biped_With_Face' preset, all dropdown should be populated correctly in all tabs. 






_____________________________________________

Get Arise at: https://www.ariserigging.com