Skip to content

Instantly share code, notes, and snippets.

@sanketgujar
Created September 3, 2018 17:28
Show Gist options
  • Save sanketgujar/58a8634cb4b29a5b112640afe8616b92 to your computer and use it in GitHub Desktop.
Save sanketgujar/58a8634cb4b29a5b112640afe8616b92 to your computer and use it in GitHub Desktop.
+---------------------------+-------------------------------------------------------------------------------+
| H function | Properties |
+---------------------------+-------------------------------------------------------------------------------+
| H(Xi,Xj) = ΘjXj | Classical Euclidean convolution |
| Θ = (Θi,....,Θk) | |
|---------------------------|-------------------------------------------------------------------------------|
| H(Xi,Xj) = H(Xi) | Encoding global information of the local neighboring structure. (Pointnet) |
|---------------------------|-------------------------------------------------------------------------------|
| H(Xi,Xj) = H(Xj - Xi) | Encodes the local information, considering the shape as collection of small |
| | patches and losing the global information. |
|---------------------------|-------------------------------------------------------------------------------|
| H(Xi,Xj) = H(Xi, Xj - Xi) | Asymmetric edge function which combines the global shape structure(Xi) and |
| | local shape features(Xj-Xi) |
+---------------------------+-------------------------------------------------------------------------------+
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment