Version: 1.2.0: 4K Ready

This update reduces memory consumption by 3x, enabling the processing of UHD/4K resolutions on GPUs with 8GB of memory.

Most optimizations were made in the Attention layer, where we consolidated inplace operations and broke down tensors into smaller batches to reduce peak memory usage.

Full Changelog: 1.1.0...1.2.0

ViTMatte for Nuke v1.1.0

🎉 First official release of ViTMatte! 🥳

Version: 1.1.0

What's New

• This release includes the complete .Cattery package for easy installation and use.
• Improved peformance and reduced memory usage with fp16 quantization.
• Added fast model. This model is more suited for improving binary masks, like those from Segment Anything for Nuke.
• Support to Nuke 13.2.
• Improved Gizmo interface
• Added overscan and bounding box support
• Fixed padding inference issues
• New toolbar icon! ✨

Installation

1. Download and unzip the latest release available here.
2. Move the extracted Cattery folder into your .nuke or your preferred location.
3. In the toolbar, select Cattery > Update or restart Nuke.

ViTMatte will then be accessible under the toolbar at Cattery > Segmentation > ViTMatte.

🐾 Extra Steps for Nuke 13

4. Add the path for ViTMatte to your init.py:

import nuke
nuke.pluginAddPath('./Cattery/vitmatte')

5. Add an menu item to the toolbar in your menu.py:

import nuke
toolbar = nuke.menu("Nodes")
toolbar.addCommand('Cattery/Segmentation/ViTMatte', 'nuke.createNode("vitmatte")', icon="vitmatte.png")

New Contributors

• @rafaelperez made their first contribution in #1
• @vinavfx made their first contribution in #4

Full Changelog: https://github.com/rafaelperez/ViTMatte-for-Nuke/commits/1.1.0