You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Currently the architecture is a modified mobilenetv3-large training on blenshapes from ARKit! The model outputs 33 blenshapes of which most cross over with SRanipals blendshapes.
5
+
6
+
### A few things to note
7
+
1. The current version of the model may not work for your mouth as it has been trained on a not very diverse dataset. Expect this to change in newer versions.
8
+
2. Not all blendshapes may work as there isnt enough data collected in the dataset.
9
+
3. This model is not a 1 to 1 translation to SRanipal blenshapes. While most blendshapes do cross over with SRanipal cheek and tounge movements are limited.
10
+
11
+
### Warranty
12
+
Lol
13
+
14
+
### Credit
15
+
16
+
Thank you to dfgHiatus#7426 for providing the unity demo as well as the intergrations with other apps!
17
+
18
+
Thank you to Rames The Generic#3540 who made the dataset, tested and evaluated the performence for the models, and for being awesome!
0 commit comments