Canvas360: Enhancing In-context Panoramic Generation via Geometric-aware Pretraining

Forum topic · 小凯 · 2026-07-12

Summary

Canvas360 is a two-stage framework for in-context panoramic generation that combines geometry-aware pretraining with downstream task-specific fine-tuning. To tackle the scarcity of large-scale, high-quality training data for in-context panorama tasks, the authors introduce Canvas360Dataset, a collection of 1 million paired panoramic samples supporting style transfer, inpainting, outpainting, and editing. On the modeling side, Canvas360 improves text-to-panorama generation through parallel depth generation, velocity circular padding, and similarity loss regularization, allowing the model to learn geometry-aware representations, capture object distortion details, and improve geometric consistency and global coherence. Leveraging strong panoramic priors, Canvas360 serves as a unified in-context panoramic generation framework that supports diverse downstream tasks via token-level concatenation, surpassing prior methods in task coverage and modeling flexibility. Experiments show improved panorama fidelity, notably strong performance on the panorama-specific FAED metric, with competitive or state-of-the-art quantitative results. Project page: https://zry000.github.io/Canvas360/

Paper Overview

Field: Computer Vision Authors: Haoran Feng, Ruiyang Zhang, Longyi Zhang, Dizhe Zhang, Lu Qi arXiv: 2607.08765 Project page: https://zry000.github.io/Canvas360/

English Abstract

In this work, the authors present Canvas360, a two-stage framework for in-context panoramic generation that combines geometry-aware pretraining with downstream task-specific fine-tuning. To address the lack of large-scale, high-quality training data tailored to in-context panoramic tasks, they propose Canvas360Dataset, a collection of 1M high-quality paired panoramic samples for style transfer, inpainting, outpainting, and editing, enabling effective supervision across diverse in-context generation scenarios. On the modeling side, Canvas360 enhances text-to-panorama generation through parallel depth generation, velocity circular padding, and similarity loss regularization, enabling the model to learn geometry-aware representations, capture object distortion details, and improve geometric consistency and global coherence. Furthermore, with strong panoramic priors, Canvas360 realizes a unified in-context panoramic generation framework that supports diverse downstream tasks via token-level concatenation, surpassing prior methods in task coverage and modeling flexibility. Extensive experiments demonstrate that Canvas360 improves panoramic image fidelity, achieving particularly strong results on the panorama-specific FAED metric, with competitive or leading performance across all reported quantitative evaluations.

Key Contributions

Canvas360Dataset: 1M high-quality paired panoramic samples covering style transfer, inpainting, outpainting, and editing tasks.
Geometry-aware modeling: parallel depth generation, velocity circular padding, and similarity loss regularization for improved geometric consistency.
Unified framework: token-level concatenation supports multiple in-context panoramic tasks beyond text-to-panorama.
Results: improved fidelity with leading panorama-specific FAED metric performance.