Even with the boiling, that's not bad at all! I would imagine you could throw it through temporalmedian and train copycat off 20-30 of the best frames n get a pretty solid result
For those having issues with the cat file error, I pulled the ViTMatte cat file from VFXEdu’s video, I paused on the GitHub frame with the google drive link displayed and typed it out in my browser (czcams.com/video/AtQv8G3jL8Q/video.htmlsi=w8C_WPFta79-x_nq) there was a link to a good drive with the matte model which is gone with this updated release. Open the vitmatte gizmo, remove the inference node, replace with a new one looking at the cat model from the google drive link, worked for me after that.
ViTMatte not working no matter what I do. Error: "You have hit the maximum amount of memory that your GPU can use. Please try reducing the size of the input image." The plate is 4k. Uncompressed exr image sequence. I have 12GB in my GPU and allocated over 8GB in Hardware Preference setting in Nuke. I have tried all Inference format settings and still the same error. I have cropped the image and reformatted to a quarter of 4k and STILL getting the same error. Followed all instructions in video and still getting this error. Nuke 15.0v4. Latest version of VitMatte installed correctly. Any assistance appreciated.
@@ytronik well, no. that won't work as the aspect ratio is not square. I have already reformatted to about that size with the correct aspect ration and it still doens't work.
Edit 02 - Works ! I forgot to put a roto node before the inference node. Cheers. End Edit 02 Edit 01 - I downloaded the folder from git and put the external cat files in the same folder. Now the node is showing up in Nuke but it's not doing anything. End Edit 01 Hi Alex . The nodes not showing up in nukex 15 . The archive which is linked to a g drive only has a cat and a pt file. Unlike the other folders that have icons and gizmo files. Did you also download the archive from the external link ? Cheers, b
@@alexvillabon Here's a small hack. I got an initial pass via the segment anything node (thanks to your video) and used that as the soft outer matte described in this video. Cheers, b
There was another person commenting with the same issue which has now been resolved. I have a feeling you didnt read the whole instructions on the github page. Take a look there.
I found the edge thickness to be not very useful tbh. Internally, it's just two FilerErodes - one expanding, one extracting the given matte, and mixing them so the larger one becomes the 0.5 the inference expects; after the inference, it gets masked by the larger one. So far, I always got better results by putting that to 0 and feeding a manually adjusted "double mask" with the "translucency range" set to 0.5
God that sounded harsh. I meant that at least I get the better results with a bit more setup, what you can pull out of the footage with only the most basic shapes is still insane.
Hey Lucas. I wasnt sure why you were annoyed at what I showed haha. Happy to hear that was not the case. Yes, what I show here just scratches the surface of what you can do. My intent with this video was to just show what you can achieve very quickly with very little effort, you can absolutely improve the results be taking a little more time and getting creative. For instance, I've had good success pairing it with copycat.
@@alexvillabon yeah it was probably a bit too late to comment on CZcams, ha! Not at all annoyed. I'm really intrigued by it; today I tested a bit more with "bad mattes", and now I got a epiphany: you can do a really really rough shape for your first slap comp, and then just improve with better input later on when there's more time or whatever. I think I also commented under your last(?) video about using it to train further with CopyCat. I think this is a really mighty combo indeed!
Works on CPU but on GPU i got this message. (Nuke14, Windows). RuntimeError: Input type (CUDAFloatType) and weight type (CUDAHalfType) should be the same Investigating what it could be...
That's strange, sounds to me like you didn't download everything. Did you read and follow all the instructions on the GitHub? I've added it to a few computers and haven't had an issue.
I downloaded the files, including the extra ones and put them in the vitmatte folder. When previewing the node, I get a really long error: Exception caught processing model: The following operation failed in the TorchScript interpreter. Traceback of TorchScript, serialized code (most recent call last): File .../__torch__.py, line 13, in forward image_and_trimap = {"image": image, "trimap": trimap} model = self.model _0 = ((model).forward(image_and_trimap, ))["phas"] ~~~~~~~~~~~~~~
Even with the boiling, that's not bad at all! I would imagine you could throw it through temporalmedian and train copycat off 20-30 of the best frames n get a pretty solid result
Absolutely!
For those having issues with the cat file error, I pulled the ViTMatte cat file from VFXEdu’s video, I paused on the GitHub frame with the google drive link displayed and typed it out in my browser (czcams.com/video/AtQv8G3jL8Q/video.htmlsi=w8C_WPFta79-x_nq) there was a link to a good drive with the matte model which is gone with this updated release. Open the vitmatte gizmo, remove the inference node, replace with a new one looking at the cat model from the google drive link, worked for me after that.
Thanks so much for this :)
sorry, where i could find and put cat file? I read gethub, but not understand((
plz installation tutorial video 😂😂😂😂😂😂😂😂😂
I wish there was a way it could output roto shapes so artists can take over and make corrections.
well technically you can try to extract alpha from a shot and then use auto-trace in after effects to extract the shapes
How to install ViTMatte in nuke 15
this is insanely good. Thanks for sharing
unfortunately I was not able to install it.I have experience in installing many cattery gizmo etc. but not this one:(
i am already using to help keying green screens an it works very well..
Indeed! Ive used it on green screens before as well.
@alexvillabon how to install ViTMatte in nuke
🤯Cool!
ViTMatte not working no matter what I do. Error: "You have hit the maximum amount of memory that your GPU can use. Please try reducing the size of the input image." The plate is 4k. Uncompressed exr image sequence. I have 12GB in my GPU and allocated over 8GB in Hardware Preference setting in Nuke. I have tried all Inference format settings and still the same error. I have cropped the image and reformatted to a quarter of 4k and STILL getting the same error. Followed all instructions in video and still getting this error. Nuke 15.0v4. Latest version of VitMatte installed correctly. Any assistance appreciated.
Just resize the input to 1024x1024
@@ytronik well, no. that won't work as the aspect ratio is not square. I have already reformatted to about that size with the correct aspect ration and it still doens't work.
@@JasonDuncan-l7s crop it, or just resize it.. if not you need more gpu
Edit 02 - Works ! I forgot to put a roto node before the inference node. Cheers. End Edit 02
Edit 01 - I downloaded the folder from git and put the external cat files in the same folder. Now the node is showing up in Nuke but it's not doing anything. End Edit 01
Hi Alex . The nodes not showing up in nukex 15 . The archive which is linked to a g drive only has a cat and a pt file. Unlike the other folders that have icons and gizmo files.
Did you also download the archive from the external link ?
Cheers,
b
Nice! happy to hear you got it working!
@@alexvillabon Here's a small hack. I got an initial pass via the segment anything node (thanks to your video) and used that as the soft outer matte described in this video.
Cheers,
b
@@behrampatel4872 smart!
Not working for me, i have other cat tools added as well but this one is saying "no cat file is selected" any solution?
There was another person commenting with the same issue which has now been resolved. I have a feeling you didnt read the whole instructions on the github page. Take a look there.
Where it is available??? Send me link
I found the edge thickness to be not very useful tbh. Internally, it's just two FilerErodes - one expanding, one extracting the given matte, and mixing them so the larger one becomes the 0.5 the inference expects; after the inference, it gets masked by the larger one. So far, I always got better results by putting that to 0 and feeding a manually adjusted "double mask" with the "translucency range" set to 0.5
God that sounded harsh. I meant that at least I get the better results with a bit more setup, what you can pull out of the footage with only the most basic shapes is still insane.
Hey Lucas. I wasnt sure why you were annoyed at what I showed haha. Happy to hear that was not the case. Yes, what I show here just scratches the surface of what you can do. My intent with this video was to just show what you can achieve very quickly with very little effort, you can absolutely improve the results be taking a little more time and getting creative. For instance, I've had good success pairing it with copycat.
@@alexvillabon yeah it was probably a bit too late to comment on CZcams, ha! Not at all annoyed. I'm really intrigued by it; today I tested a bit more with "bad mattes", and now I got a epiphany: you can do a really really rough shape for your first slap comp, and then just improve with better input later on when there's more time or whatever.
I think I also commented under your last(?) video about using it to train further with CopyCat. I think this is a really mighty combo indeed!
cool
Am I the only one who cant figure out how to download it?) Link in description lead to "There aren’t any releases here"
Well. Now I know, but it is quite tricky)
All the info is on the github. Dont just jump to releases.
@@AlexeyKorotkikh hey, did you find a way to download the Cattery folder?
@@sen73nced Yes, there is no realize zip. You need to download each file separately from folder icon above description
thank you, great tool!! And I want to ask you Which tool is 10.39 ?
That would be w_hotbox. You can grab it here www.nukepedia.com/python/ui/w_hotbox/
@@alexvillabon thank you
Works on CPU but on GPU i got this message. (Nuke14, Windows).
RuntimeError: Input type (CUDAFloatType) and weight type (CUDAHalfType) should be the same
Investigating what it could be...
Works now if I reformat to a square frame
Great!
Hi thank you for the vid, followed the instruction but when adding the node I get an error 'no cat file is selected' any idea?
That's strange, sounds to me like you didn't download everything. Did you read and follow all the instructions on the GitHub? I've added it to a few computers and haven't had an issue.
@@alexvillabon yes you are correct i forgot the cat file :) thanks!
@@kobiohanna great!
I downloaded the files, including the extra ones and put them in the vitmatte folder. When previewing the node, I get a really long error:
Exception caught processing model: The following operation failed in the TorchScript interpreter.
Traceback of TorchScript, serialized code (most recent call last):
File .../__torch__.py, line 13, in forward
image_and_trimap = {"image": image, "trimap": trimap}
model = self.model
_0 = ((model).forward(image_and_trimap, ))["phas"]
~~~~~~~~~~~~~~
Ahh, okay apparently it was similar to somebody else's problem. I just had to reformat because my pixel aspect was 1.5 on this footage. Thanks!
Great!
@@alexvillabon I have the same error. How to fix it?