OpenCubeDiff
Collection
Open-source implementation of "CubeDiff: Repurposing Diffusion-Based Image Models for Panorama Generation". • 3 items • Updated
import torch
from diffusers import DiffusionPipeline
# switch to "mps" for apple devices
pipe = DiffusionPipeline.from_pretrained("hlicai/cubediff-512-multitxt", dtype=torch.bfloat16, device_map="cuda")
prompt = "Astronaut in a jungle, cold color palette, muted colors, detailed, 8k"
image = pipe(prompt).images[0]CubeDiff Panorama Generation Model - Multi Text
This is an open-source implementation of CubeDiff, a method for 360° panorama generation based on diffusion models.
Please refer to the official paper and project page for more information:
📄 Paper: CubeDiff: Repurposing Diffusion-Based Image Models for Panorama Generation
🌐 Original Project Page: Cubediff
📚 Open-source implementation: OpenCubeDiff
Model Details
This model is part of the CubeDiff open-source reimplementation, carried out as part of a semester project by
Hanqiu Li Cai and Juan Tarazona Rodríguez for their Master's degree in Robotics, Systems and Control at ETH Zürich.
This model is not affiliated in any shape or form with Google.
We repurpose and fine-tune a Stable Diffusion backbone (SD 1.5) to generate cube-face-consistent panoramas using CubeDiff-style attention reshaping and conditioning.
For installation, usage examples, and training details, please visit the project repository:
🔗 https://github.com/Juan5713/OpenCubeDiff
- Downloads last month
- 82
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support