Skeleton design

Contents

Skeleton design#

In SLEAP, skeletons are defined as a set of nodes (landmarks types) and edges (connections between landmarks).

Nodes are essentially just a list of body part names. The actual naming doesn’t matter; SLEAP just uses their relative order to keep track of which body part is which in the labels and in trained models.

Edges describe the way that the nodes are connected. Each edge is represented by a source and destination node. With the exception of bottom-up models, these edges serve primarily for visualization.

In bottom-up models, the edges are important and must connect all nodes to one another.

Tips#

How do I choose the best skeleton for my data?

In general, here are the rules of thumb you want to follow for designing an ideal skeleton:

  1. Choose nodes that will be easy to locate in new images. It’s important to be as consistent as possible about the relative placement of body parts.

  2. When choosing edges, try to form a shallow tree. This is because it’s preferable to have fewer parent nodes, since if they are not detected, their children nodes cannot be grouped appropriately.

Can I add or remove nodes in the skeleton after I’ve already created instances?

Yes.

Removing nodes is straightforward: just remove them. If you’re using part affinity fields for inference, you should make sure that the skeleton graph is still connected.

Adding nodes is a little more complicated. First add the nodes to your skeleton. Then, to add these nodes to any instance which already exists, you’ll need to double-click on the instance (on the video frame image). The new nodes will be added and marked as “non-visible”; you’ll need to right-click on each node you want to make visible, and move it to the correct location.

Can I add or remove edges in the skeleton after I’ve already created instances?

Yes, adding or removed edges is straightforward and the change will be applied to all instances.