Implement a setting to average left/right face expressions#194
Implement a setting to average left/right face expressions#194Erimelowo wants to merge 7 commits intoProject-Babble:mainfrom
Conversation
|
Hey Erimelowo, thanks for your PR. I am having a look over it now - to be clear, the way this setting works is it biases an expression towards a left or right, given some factor above? IE 0.5 is no bias, 0 is left and right is 1 |
0 is vanilla behaviour, 0.5 is so each side interpolates 50% towards the average of both sides and 100% makes it so both sides are equal to the average of both sides. Example, if left is 0.2 and right is 0.6, that means the average is 0.4 and that |
|
im still unsure about this change as it significantly changes behavior of the output. we need to do more testing to verify that this is an improvement with our larger userbase |
|
It is disabled by default… so far my tracking has been great (much better than before) with this feature. |
|
Hey there, I've made it home and checked out this PR. Following up on what Summer said I have some usability concerns - while this is opt in, I worry it may not be immediately apparent to end users what and how this feature works. |
|
I don't want to egg things on. I'm going to close this PR until we have a better means of conveying to end users how this feature is used. |
|
Hi! As an end user who has a similar issue this PR would solve, I'd love to see this reopened! It seems clear to me what it does, comparable to existing settings like the "One-Euro filter settings". If testing is what's blocking this from moving forward, I'd be happy to run it for a bit on my end and report back c: |
|
Hello, I can take a look at this again. Additionally, I can probably rename the one euro filter setting to something more friendly |



Adds a section under Stabilize Eyes that adds a setting to average left and right expressions according to certain percentages. This can increase stability at the cost of fidelity.
I struggled a bit with making it affect the preview and decided it wasn't worth it since it might make further calibration harder. I could try my hand at it again if that preview should match the tracking.
Closes #193