Tomorrow Today Last Wednesday, on the 13th of May, the manual Patrons group (the one that gives you your green username AND access to dedicated Patrons spaces on WSL) will be has been DELETED.
This means that your access to those spaces will be has been revoked until you connect your Patreon account to WSL using the link below, which will give you access to the NEW, future-proof Patrons+++ group.
With the imminent release of ZEPPO BETA 2 - tentatively planned for released Friday today yesterday (!!) - I wouldn't dawdle... with Beta 2 (i.e. all my late nights in the past month), Zeppo has become something truly special, battle-tested and multi-user hardened in an actual production.
Yes, there will be a Reactor release eventually, and no, it won't have all the bells and whistles - those are for Patrons, because they are the ones keeping WSL afloat, which is still only barely these days...
I'd better get back to prepping that done did release then!
Thank you,
Pieter
Where the future is being made, today.
Yes, leave it to me to uncover a completely unrelated bug while testing someone else's tool!
What method did you use? I think you mentioned Claude to work stuff out. I still think the fSpy rectangle method might be the cleanest approach. And since it's both open source and AFAIK not dependent on an external library like openCV it might be the most accessible method?
I have the feeling that most of the stuff in this document could be used for the perspective crop part. Especially the stuff around page 13. Section 4.3
v0.5 2025-03-25:
@ShadowMaker SdR Apparently the 3Rn thing isn't a bug. Fusion have a DoD and it has a DataWindow. The data window is what contains the data itself while DoD shows the valid window of the image. DCTL reads the DataWindows content from 0-1 so if the DoD is filling the whole image the DCTL kernel won't read the empty data between the DoD and the data window. 3Rn is the only place I have come across where the data window isn't the same as the valid window, but now we know 
In short, I had to rewrite all DoD calculations but now it works as expected!
Maybe it's time to create a Fuse cource with all of these kind-of-edge-cases-but-not-really things 
v0.6 2025-03-26:
v0.7 2025-04-01:
v0.8 2025-04-13:
I think I finally got the RoI request working correctly. I thought it was smiple but man, it required quite some math 
I also got tracker-linking working. Simply put a tracker node in the Tracker box and press Connect. Everything will then be connected for you.
Not tho that you currently can't pick what tracker goes to what corner.
I will need to figure out how to automatically populate and update dropdown boxes for this to work. I figured out how to add entries but I coulnd't get the list to update. The user would need to select another node and then back to ConerPinner to see the new entries. I also have no idea how to remove entries.
Was any of the info I mentioned useful for the aspect ratio calculation? I am still convinced that if we could figure out the method fSpy uses (and it's open source, so it should be doable to find it) then everything would be fine. I've fed it my test images and the rectangle method it uses to calculate the complete camera matches 1:1 with the image I'm feeding it, when reconstructing the geometry in Blender.
There is some parts that needs some updating after all features is in. The perspective mode will absolutely need to be re-written. It will come
And with that we're on 0.9!
v0.9 2025-04-13:
So the only thing truly left is to fix the perspective crop After that it's cleanup time!
v0.10 2025-04-14:
@ShadowMaker SdR I did my best to get as good Perspective Crop as possible. I tried to understand the things you sent but man, when math becomes more letter than numbers I'm out 
I had to Claude a lot, tried different algorithms etc and this is the best one I could come up with.
I tried to see where in fSpy's code the perspective calculation happens but coulnd't pin point it.
If you can find it you're more than welcome to try and add it to the fuse!
A bit late here but version 0.10, at the bottom of the inspector, still refers to it as 0.9.
