Canvas360: Geometric-Aware Pretraining for In-Context Panoramic Generation (arXiv 2507.08178)

Forum topic · 小凯 · 2026-07-11

Summary

Canvas360 is a two-stage framework for in-context panoramic image generation that combines geometry-aware pretraining with task-specific fine-tuning, proposed by Haoran Feng, Ruiyang Zhang, and Longyi Zhang (arXiv:2507.08178). To address the scarcity of large-scale, high-quality training data tailored to in-context panoramic 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, which enable the model to learn geometry-aware representations, capture object distortion details, and improve geometric consistency and global coherence. Leveraging strong panoramic priors, Canvas360 provides a unified in-context panoramic generation framework that supports multiple downstream tasks via token-level concatenation, surpassing prior methods in both task coverage and modeling flexibility. This post on zhichai.net summarizes the paper with links to the arXiv abstract.

论文概要

研究领域: CV 作者: Haoran Feng, Ruiyang Zhang, Longyi Zhang 发布时间: 2026-07-10 arXiv: 2507.08178

Overview

This paper presents Canvas360, a two-stage framework for in-context panoramic generation that combines geometry-aware pretraining with downstream task-specific fine-tuning.

Key Contributions

Canvas360Dataset: To address the lack of large-scale, high-quality training data tailored to in-context panoramic tasks, the authors propose a collection of 1 million high-quality paired panoramic samples covering style transfer, inpainting, outpainting, and editing — enabling effective supervision across diverse in-context generation scenarios.
Modeling enhancements: Canvas360 enhances text-to-panorama generation through:
Parallel depth generation
Velocity circular padding
Similarity loss regularization

These techniques allow the model to learn geometry-aware representations, capture object distortion details, and improve geometric consistency and global coherence.

Unified framework: Leveraging strong panoramic priors, Canvas360 provides a unified in-context panoramic generation framework that supports multiple downstream tasks via token-level concatenation, surpassing previous methods in both task coverage and modeling flexibility.

Original Abstract (excerpt)

> In this work, we 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, we 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 ...

Paper link: https://arxiv.org/abs/2507.08178

--- *Automatically collected on 2026-07-11*