pith. sign in

arxiv: 2606.27745 · v1 · pith:QDEB3O57new · submitted 2026-06-26 · 💻 cs.CV

Panoramic Scene Analysis: A Survey from Distortion-Aware Engineering to Sphere-Native Foundation Modeling

Qinfeng Zhu , Lei Fan This is my paper

Pith reviewed 2026-06-29 04:33 UTC · model grok-4.3

classification 💻 cs.CV
keywords panoramic scene analysisspherical equivariancedistortion-aware engineeringsphere-native modelingfoundation modelsevaluation protocolsgeometric adaptation360-degree vision
0
0 comments X

The pith

No panoramic scene analysis method achieves both strict spherical equivariance and full reuse of perspective-pretrained foundation-model weights.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

This survey traces how panoramic scene analysis has moved from projection-based fixes and distortion-aware tweaks to sphere-native architectures and geometry-aware tokenization. It organizes the field along two axes: how network operators respect spherical geometry and how training reuses knowledge from flat-image models. A central finding is that these two requirements remain in tension across all covered techniques for tasks like segmentation, depth estimation, layout prediction, and vision-language reasoning. The paper also flags five missing elements in current evaluation practices and sketches a roadmap for closing the gaps toward general panoramic intelligence.

Core claim

The literature evolves along a trajectory of increasing geometric commitment from projection adaptation through distortion engineering to sphere-native modeling, yet none of the surveyed methods simultaneously delivers strict spherical equivariance and full reuse of perspective-pretrained foundation-model weights; this tension is structural. Five systematic gaps exist in evaluation protocols: absence of spherical-area-weighted metrics, seam-consistency testing, polar-robustness stratification, cross-projection generalization checks, and open-world protocol standardization. A six-point research roadmap is proposed to reach general-purpose panoramic intelligence.

What carries the argument

The two orthogonal dimensions of architectural design (operator interaction with spherical geometry) and training paradigm (knowledge transfer across domains) that together expose the unresolved equivariance-versus-weight-reuse tension.

If this is right

  • Architectures must be redesigned so spherical geometry is native rather than corrected after the fact.
  • Training paradigms need mechanisms that preserve pretrained weights while enforcing exact equivariance under spherical transformations.
  • Evaluation protocols must add area-weighted metrics, seam-consistency tests, polar stratification, cross-projection checks, and standardized open-world benchmarks.
  • Multi-task, open-world, and video panoramic systems will remain limited until the equivariance-weight tension is resolved.
  • Geometry-aware tokenization offers one concrete route toward unified panoramic foundation models.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • A hybrid approach that freezes most pretrained weights and adds lightweight spherical-equivariant layers could serve as an immediate engineering bridge.
  • The same tension likely appears in other omnidirectional sensing domains such as 360 video compression or robotic navigation on curved surfaces.
  • If the gap is truly structural, native spherical pretraining from scratch on large panoramic corpora may ultimately outperform any adaptation strategy.
  • Standard vision benchmarks could be extended with synthetic spherical distortions to quantify how much equivariance is lost when reusing flat-image weights.

Load-bearing premise

The surveyed methods comprehensively represent the current state of the art, so the identified tension between equivariance and weight reuse is genuinely unresolved rather than merely overlooked.

What would settle it

Publication or discovery of any single method that simultaneously achieves strict spherical equivariance and full reuse of perspective-pretrained foundation-model weights on the covered tasks would falsify the structural-gap claim.

Figures

Figures reproduced from arXiv: 2606.27745 by Lei Fan, Qinfeng Zhu.

Figure 1
Figure 1. Figure 1: Panoramic imaging pipeline. Diverse capture systems (a) produce signals on the unit sphere (b), which must be projected to [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Two-dimensional taxonomy of panoramic modeling. The horizontal axis represents architectural sophistication in spherical [PITH_FULL_IMAGE:figures/full_fig_p007_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Four structural challenges that distinguish dynamic panoramic perception from perspective video. (a) Identical 3D motion [PITH_FULL_IMAGE:figures/full_fig_p023_3.png] view at source ↗
read the original abstract

Panoramic images capture the complete visual sphere in a single frame, providing spatial context unattainable by conventional cameras. Yet this completeness comes at a geometric cost: the 2-sphere cannot be faithfully mapped to the plane, and every planar representation introduces distortions that violate the assumptions underlying standard vision architectures. This survey traces the evolution of panoramic scene analysis along a methodological trajectory, from projection-based adaptation, through distortion-aware engineering, to sphere-native modeling and geometry-aware tokenization for foundation models, and argues that this evolution reflects a progressive deepening of geometric commitment rather than a simple accumulation of techniques. We organize the literature along two orthogonal dimensions: architectural design (how operators interact with spherical geometry) and training paradigm (how knowledge is transferred across domains). Covering dense prediction (semantic segmentation, depth estimation, and room layout estimation), unified multi-task understanding, open-world perception, vision-language reasoning, and dynamic video analysis, we identify a central unresolved tension: among the methods surveyed, none simultaneously delivers strict spherical equivariance and full reuse of perspective-pretrained foundation-model weights, and we argue that this is a structural rather than incidental gap. We further expose five systematic gaps in current evaluation protocols, namely the absence of spherical-area-weighted metrics, seam-consistency testing, polar-robustness stratification, cross-projection generalization, and open-world protocol standardization, and propose a six-point research roadmap toward general-purpose panoramic intelligence. The corresponding repository is publicly available at: https://github.com/zhuqinfeng1999/Awesome-Panoramic-Scene-Analysis.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. This survey traces the development of panoramic scene analysis from projection-based adaptations and distortion-aware engineering to sphere-native modeling and geometry-aware tokenization. It organizes the literature along two axes—architectural design (operator interaction with spherical geometry) and training paradigm (knowledge transfer)—covering tasks from dense prediction and multi-task understanding to vision-language reasoning and video analysis. The central claim is that no surveyed method simultaneously achieves strict spherical equivariance and full reuse of perspective-pretrained foundation-model weights, which the authors argue is a structural gap; the paper also identifies five evaluation-protocol gaps and outlines a six-point research roadmap, with an accompanying public repository.

Significance. If the survey coverage is representative, the work would usefully synthesize progress in panoramic vision and isolate a concrete tension between geometric fidelity and transfer learning that future foundation-model efforts must resolve. The explicit identification of evaluation gaps (spherical-area-weighted metrics, seam consistency, polar robustness, cross-projection generalization, open-world protocols) and the public repository provide concrete value for the community even if the structural-gap claim requires further substantiation.

major comments (2)
  1. [Abstract and §1] Abstract and §1 (Introduction): The claim that 'none simultaneously delivers strict spherical equivariance and full reuse of perspective-pretrained foundation-model weights' and that the tension is 'structural rather than incidental' is the paper's primary contribution. This assertion rests on the surveyed set being exhaustive, yet the manuscript provides no documented search protocol, keyword list, database sources, date cutoffs, or explicit inclusion/exclusion rules, preventing verification that omitted recent adapter-based or hybrid equivariant fine-tuning approaches do not already satisfy both criteria.
  2. [§4 and §5] §4 (Training Paradigm) and §5 (Evaluation Gaps): The five listed evaluation gaps are presented as systematic, but the section does not quantify how many of the surveyed papers fail each criterion or provide a table mapping methods to the gaps; without this, it is unclear whether the gaps are uniformly unaddressed or whether a subset of methods already partially satisfies them, weakening the roadmap's motivation.
minor comments (2)
  1. The two-dimensional taxonomy (architectural design × training paradigm) is conceptually clean, but the manuscript would benefit from an explicit table or figure that places every cited method into the taxonomy cells to improve traceability.
  2. Several section headings use similar phrasing (e.g., 'sphere-native' appears in multiple contexts); consistent terminology or a glossary would reduce ambiguity for readers new to the subfield.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their insightful comments, which help improve the rigor of our survey. We provide point-by-point responses to the major comments below, indicating planned revisions to address the concerns about documentation and substantiation of our claims.

read point-by-point responses

  1. Referee: [Abstract and §1] Abstract and §1 (Introduction): The claim that 'none simultaneously delivers strict spherical equivariance and full reuse of perspective-pretrained foundation-model weights' and that the tension is 'structural rather than incidental' is the paper's primary contribution. This assertion rests on the surveyed set being exhaustive, yet the manuscript provides no documented search protocol, keyword list, database sources, date cutoffs, or explicit inclusion/exclusion rules, preventing verification that omitted recent adapter-based or hybrid equivariant fine-tuning approaches do not already satisfy both criteria.

    Authors: We agree that a documented search protocol is essential for verifying the exhaustiveness of the surveyed literature and thus the strength of our central claim. In the revised version, we will insert a new subsection (e.g., §2.1 Literature Search Methodology) that explicitly details the databases queried (Google Scholar, arXiv, IEEE Xplore), the keyword combinations employed, the time period covered, and the inclusion/exclusion criteria applied. This addition will enable independent verification and address potential concerns about omitted methods such as recent adapter-based approaches. revision: yes

  2. Referee: [§4 and §5] §4 (Training Paradigm) and §5 (Evaluation Gaps): The five listed evaluation gaps are presented as systematic, but the section does not quantify how many of the surveyed papers fail each criterion or provide a table mapping methods to the gaps; without this, it is unclear whether the gaps are uniformly unaddressed or whether a subset of methods already partially satisfies them, weakening the roadmap's motivation.

    Authors: We concur that quantifying the prevalence of each gap and providing a mapping would strengthen the motivation for the proposed roadmap. Accordingly, we will augment §5 with a comprehensive table that lists all surveyed methods and indicates their compliance with each of the five evaluation criteria. We will also include summary statistics (e.g., percentage of methods addressing each gap) to demonstrate that the gaps are indeed widespread. This revision will clarify the systematic nature of the identified issues. revision: yes

Circularity Check

0 steps flagged

No circularity: survey paper with no derivations or self-referential fits

full rationale

This is a literature survey paper with no equations, fitted parameters, predictions, or derivation chains. The central claim—that no surveyed method achieves both strict spherical equivariance and full weight reuse—rests on external reviewed publications rather than any reduction to the paper's own inputs or self-citations. No self-definitional, fitted-input, or uniqueness-imported patterns apply. The paper is self-contained as a review whose exhaustiveness can be externally checked against the cited works.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The survey rests on the domain assumption that the reviewed body of work is representative; no free parameters or invented entities are introduced.

axioms (1)
  • domain assumption The surveyed literature is representative of the field and sufficient to establish that no method achieves both strict spherical equivariance and full weight reuse.
    The central tension claim depends on exhaustive coverage of relevant methods.

pith-pipeline@v0.9.1-grok · 5814 in / 1234 out tokens · 58326 ms · 2026-06-29T04:33:07.346335+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

145 extracted references · 40 canonical work pages · 10 internal anchors

  1. [1]

    Hao Ai, Zidong Cao, Yan-Pei Cao, Ying Shan, and Lin Wang. 2023. Hrdfuse: Monocular 360deg depth estimation by collaboratively learning holistic-with-regional depth distributions. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 13273–13282

    2023

  2. [2]

    Hao Ai, Zidong Cao, Jinjing Zhu, Haotian Bai, Yucheng Chen, and Lin Wang. 2022. Deep learning for omnidirectional vision: A survey and new perspectives.arXiv preprint arXiv:2205.10468(2022)

    work page arXiv 2022

  3. [3]

    Hao Ai and Lin Wang. 2024. Elite360d: Towards efficient 360 depth estimation via semantic-and distance-aware bi-projection fusion. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 9926–9935

    2024

  4. [4]

    Hao Ai and Lin Wang. 2024. Elite360m: Efficient 360 multi-task learning via bi-projection fusion and cross-task collaboration.arXiv preprint arXiv:2408.09336(2024)

    work page arXiv 2024

  5. [5]

    Georgios Albanis, Nikolaos Zioulis, Petros Drakoulis, Vasileios Gkitsas, Vladimiros Sterzentsenko, Federico Alvarez, Dimitrios Zarpalas, and Petros Daras. 2021. Pano3d: A holistic benchmark and a solid baseline for 360deg depth estimation. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 3727–3737

    2021

  6. [6]

    Peter Anderson, Qi Wu, Damien Teney, Jake Bruce, Mark Johnson, Niko Sünderhauf, Ian Reid, Stephen Gould, and Anton Van Den Hengel. 2018. Vision-and-language navigation: Interpreting visually-grounded navigation instructions in real environments. InProceedings of the IEEE conference on computer vision and pattern recognition. 3674–3683

    2018

  7. [7]

    Iro Armeni, Sasha Sax, Amir R Zamir, and Silvio Savarese. 2017. Joint 2d-3d-semantic data for indoor scene understanding.arXiv preprint arXiv:1702.01105(2017)

    work page internal anchor Pith review Pith/arXiv arXiv 2017

  8. [8]

    Jiayang Bai, Haoyu Qin, Shuichang Lai, Jie Guo, and Yanwen Guo. 2024. GLPanoDepth: Global-to-local panoramic depth estimation.IEEE Transactions on Image Processing33 (2024), 2936–2949

    2024

  9. [9]

    Yaniv Benny and Lior Wolf. 2025. Sphereuformer: A u-shaped transformer for spherical 360 perception. InProceedings of the Computer Vision and Pattern Recognition Conference. 940–950

    2025

  10. [10]

    Meiqi Cao, Rui Yan, Xiangbo Shu, Guangzhao Dai, Yazhou Yao, and Guo-Sen Xie. 2024. Adafpp: Adapt-focused bi-propagating prototype learning for panoramic activity recognition. InProceedings of the 32nd ACM International Conference on Multimedia. 691–700

    2024

  11. [11]

    Yihong Cao, Jiaming Zhang, Hao Shi, Kunyu Peng, Yuhongxuan Zhang, Hui Zhang, Rainer Stiefelhagen, and Kailun Yang. 2024. Occlusion-aware seamless segmentation. InEuropean Conference on Computer Vision. Springer, 129–147

    2024

  12. [12]

    Zidong Cao, Jinjing Zhu, Weiming Zhang, Hao Ai, Haotian Bai, Hengshuang Zhao, and Lin Wang. 2025. Panda: Towards panoramic depth anything with unlabeled panoramas and mobius spatial augmentation. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 982–992

    2025

  13. [13]

    Mahdi Chamseddine, Didier Stricker, and Jason Rambach. 2026. PanoSAMic: Panoramic Image Segmentation from SAM Feature Encoding and Dual View Fusion.arXiv preprint arXiv:2601.07447(2026)

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  14. [14]

    Angel Chang, Angela Dai, Thomas Funkhouser, Maciej Halber, Matthias Niessner, Manolis Savva, Shuran Song, Andy Zeng, and Yinda Zhang

  15. [15]

    Matterport3d: Learning from rgb-d data in indoor environments.arXiv preprint arXiv:1709.06158(2017)

    work page internal anchor Pith review Pith/arXiv arXiv 2017

  16. [16]

    Boyuan Chen, Zhuo Xu, Sean Kirmani, Brain Ichter, Dorsa Sadigh, Leonidas Guibas, and Fei Xia. 2024. Spatialvlm: Endowing vision-language models with spatial reasoning capabilities. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 14455–14465

    2024

  17. [17]

    Hsien-Tzu Cheng, Chun-Hung Chao, Jin-Dong Dong, Hao-Kai Wen, Tyng-Luh Liu, and Min Sun. 2018. Cube padding for weakly-supervised saliency prediction in 360 videos. InProceedings of the IEEE conference on computer vision and pattern recognition. 1420–1429

    2018

  18. [18]

    Shih-Han Chou, Wei-Lun Chao, Wei-Sheng Lai, Min Sun, and Ming-Hsuan Yang. 2020. Visual question answering on 360deg images. InProceedings of the IEEE/CVF winter conference on applications of computer vision. 1607–1616. Manuscript submitted to ACM 32 Zhu and Fan

    2020

  19. [19]

    Taco Cohen, Maurice Weiler, Berkay Kicanaoglu, and Max Welling. 2019. Gauge equivariant convolutional networks and the icosahedral CNN. In International conference on Machine learning. PMLR, 1321–1330

    2019

  20. [20]

    Taco S Cohen, Mario Geiger, Jonas Köhler, and Max Welling. 2018. Spherical cnns.arXiv preprint arXiv:1801.10130(2018)

    work page internal anchor Pith review Pith/arXiv arXiv 2018

  21. [21]

    Benjamin Coors, Alexandru Paul Condurache, and Andreas Geiger. 2018. Spherenet: Learning spherical representations for detection and classification in omnidirectional images. InProceedings of the European conference on computer vision (ECCV). 518–533

    2018

  22. [22]

    Marius Cordts, Mohamed Omran, Sebastian Ramos, Timo Rehfeld, Markus Enzweiler, Rodrigo Benenson, Uwe Franke, Stefan Roth, and Bernt Schiele. 2016. The cityscapes dataset for semantic urban scene understanding. InProceedings of the IEEE conference on computer vision and pattern recognition. 3213–3223

    2016

  23. [23]

    Thiago LT Da Silveira, Paulo GL Pinto, Jeffri Murrugarra-Llerena, and Cláudio R Jung. 2022. 3d scene geometry estimation from 360 imagery: A survey.Comput. Surveys55, 4 (2022), 1–39

    2022

  24. [24]

    Michaël Defferrard, Martino Milani, Frédérick Gusset, and Nathanaël Perraudin. 2020. DeepSphere: a graph-based spherical CNN.arXiv preprint arXiv:2012.15000(2020)

    work page arXiv 2020

  25. [25]

    Zihao Dongfang, Xu Zheng, Ziqiao Weng, Yuanhuiyi Lyu, Danda Pani Paudel, Luc Van Gool, Kailun Yang, and Xuming Hu. 2025. Are multimodal large language models ready for omnidirectional spatial reasoning?arXiv preprint arXiv:2505.11907(2025)

    work page arXiv 2025

  26. [26]

    Mengfei Duan, Yuheng Zhang, Yihong Cao, Fei Teng, Kai Luo, Jiaming Zhang, Kailun Yang, and Zhiyong Li. 2025. Panoramic out-of-distribution segmentation.arXiv preprint arXiv:2505.03539(2025)

    work page arXiv 2025

  27. [27]

    Marc Eder, Mykhailo Shvets, John Lim, and Jan-Michael Frahm. 2020. Tangent images for mitigating spherical distortion. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 12426–12434

    2020

  28. [28]

    Carlos Esteves, Christine Allen-Blanchette, Ameesh Makadia, and Kostas Daniilidis. 2018. Learning so (3) equivariant representations with spherical cnns. InProceedings of the european conference on computer vision (ECCV). 52–68

    2018

  29. [29]

    Carlos Esteves, Ameesh Makadia, and Kostas Daniilidis. 2020. Spin-weighted spherical cnns. InAdvances in Neural Information Processing Systems, Vol. 33. 8614–8625

    2020

  30. [30]

    Carlos Esteves, Jean-Jacques Slotine, and Ameesh Makadia. 2023. Scaling spherical cnns.arXiv preprint arXiv:2306.05420(2023)

    work page arXiv 2023

  31. [31]

    Weijia Fan, Ruiping Liu, Jiale Wei, Yufan Chen, Junwei Zheng, Zichao Zeng, Jiaming Zhang, Qiufu Li, Linlin Shen, and Rainer Stiefelhagen. 2026. More than the Sum: Panorama-Language Models for Adverse Omni-Scenes.arXiv preprint arXiv:2603.09573(2026)

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  32. [32]

    Shaohua Gao, Kailun Yang, Hao Shi, Kaiwei Wang, and Jian Bai. 2022. Review on panoramic imaging and its applications in scene understanding. IEEE Transactions on Instrumentation and Measurement71 (2022), 1–34

    2022

  33. [33]

    Jan Gerken, Oscar Carlsson, Hampus Linander, Fredrik Ohlsson, Christoffer Petersson, and Daniel Persson. 2022. Equivariance versus augmentation for spherical images. InInternational Conference on Machine Learning. PMLR, 7404–7421

    2022

  34. [34]

    Christopher Geyer and Kostas Daniilidis. 2000. A unifying theory for central panoramic systems and practical implications. InEuropean conference on computer vision. Springer, 445–461

    2000

  35. [35]

    Krzysztof M Gorski, Eric Hivon, Anthony J Banday, Benjamin D Wandelt, Frode K Hansen, Mstvos Reinecke, and Matthia Bartelmann. 2005. HEALPix: A framework for high-resolution discretization and fast analysis of data distributed on the sphere.The Astrophysical Journal622, 2 (2005), 759–771

    2005

  36. [36]

    Yuliang Guo, Sparsh Garg, S Mahdi H Miangoleh, Xinyu Huang, and Liu Ren. 2025. Depth any camera: Zero-shot metric depth estimation from any camera. InProceedings of the Computer Vision and Pattern Recognition Conference. 26996–27006

    2025

  37. [37]

    Ruize Han, Haomin Yan, Jiacheng Li, Songmiao Wang, Wei Feng, and Song Wang. 2022. Panoramic human activity recognition. InEuropean Conference on Computer Vision. Springer, 244–261

    2022

  38. [38]

    Byeongho Heo, Song Park, Dongyoon Han, and Sangdoo Yun. 2024. Rotary position embedding for vision transformer. InEuropean Conference on Computer Vision. Springer, 289–305

    2024

  39. [39]

    Jie Hu, Junwei Zheng, Jiale Wei, Jiaming Zhang, and Rainer Stiefelhagen. 2024. Deformable mamba for wide field of view segmentation.arXiv preprint arXiv:2411.16481(2024)

    work page arXiv 2024

  40. [40]

    Huajian Huang, Yinzhe Xu, Yingshu Chen, and Sai-Kit Yeung. 2023. 360vot: A new benchmark dataset for omnidirectional visual object tracking. InProceedings of the IEEE/CVF International Conference on Computer Vision. 20566–20576

    2023

  41. [41]

    Kun Huang, Fanglue Zhang, and Neil Dodgson. 2024. PanoNormal: Monocular Indoor 360 ◦ Surface Normal Estimation.arXiv preprint arXiv:2405.18745(2024)

    work page arXiv 2024

  42. [42]

    Sandeep Inuganti, Hideaki Kanayama, Kanta Shimizu, Mahdi Chamseddine, Soichiro Yokota, Didier Stricker, and Jason Rambach. 2026. JOPP-3D: Joint Open Vocabulary Semantic Segmentation on Point Clouds and Panoramas.arXiv preprint arXiv:2603.06168(2026)

    work page internal anchor Pith review arXiv 2026

  43. [43]

    Md Amirul Islam, Sen Jia, and Neil DB Bruce. 2020. How much position information do convolutional neural networks encode?arXiv preprint arXiv:2001.08248(2020)

    work page arXiv 2020

  44. [44]

    Alexander Jaus, Kailun Yang, and Rainer Stiefelhagen. 2023. Panoramic panoptic segmentation: Insights into surrounding parsing for mobile agents via unsupervised contrastive learning.IEEE Transactions on Intelligent Transportation Systems24, 4 (2023), 4438–4453

    2023

  45. [45]

    Hualie Jiang, Zhe Sheng, Siyu Zhu, Zilong Dong, and Rui Huang. 2021. Unifuse: Unidirectional fusion for 360 panorama depth estimation.IEEE Robotics and Automation Letters6, 2 (2021), 1519–1526

    2021

  46. [46]

    Hualie Jiang, Ziyang Song, Zhiqiang Lou, Rui Xu, and Minglang Tan. 2025. Depth Anything in360 ◦ : Towards Scale Invariance in the Wild.arXiv preprint arXiv:2512.22819(2025). Manuscript submitted to ACM Panoramic Scene Analysis: A Survey 33

    work page arXiv 2025

  47. [47]

    Jing Jiang, Sicheng Zhao, Jiankun Zhu, Wenbo Tang, Zhaopan Xu, Jidong Yang, Guoping Liu, Tengfei Xing, Pengfei Xu, and Hongxun Yao. 2025. Multi-source domain adaptation for panoramic semantic segmentation.Information Fusion117 (2025), 102909

    2025

  48. [48]

    Lutao Jiang, Zidong Cao, Weikai Chen, Xu Zheng, Yuanhuiyi Lyu, Zhenyang Li, Zeyu Hu, Yingda Yin, Keyang Luo, Runze Zhang, et al. 2026. SAP: Segment Any 4K Panorama.arXiv preprint arXiv:2603.12759(2026)

    work page arXiv 2026

  49. [49]

    Zhigang Jiang, Zhongzheng Xiang, Jinhua Xu, and Ming Zhao. 2022. Lgt-net: Indoor panoramic room layout estimation with geometry-aware transformer network. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 1654–1663

    2022

  50. [50]

    Seongmin Jung, Seongho Choi, Gunwoo Jeon, Minsu Cho, and Jongwoo Lim. 2025. PanoGrounder: Bridging 2D and 3D with Panoramic Scene Representations for VLM-based 3D Visual Grounding.arXiv preprint arXiv:2512.20907(2025)

    work page arXiv 2025

  51. [51]

    Juho Kannala and Sami S Brandt. 2006. A generic camera model and calibration method for conventional, wide-angle, and fish-eye lenses.IEEE transactions on pattern analysis and machine intelligence28, 8 (2006), 1335–1340

    2006

  52. [52]

    Bogdan Khomutenko, Gaëtan Garcia, and Philippe Martinet. 2015. An enhanced unified camera model.IEEE Robotics and Automation Letters1, 1 (2015), 137–144

    2015

  53. [53]

    Risi Kondor and Shubhendu Trivedi. 2018. On the generalization of equivariance and convolution in neural networks to the action of compact groups. InInternational conference on machine learning. PMLR, 2747–2755

    2018

  54. [54]

    Duy Tho Le, Chenhui Gou, Stavya Datta, Hengcan Shi, Ian Reid, Jianfei Cai, and Hamid Rezatofighi. 2024. Jrdb-panotrack: An open-world panoptic segmentation and tracking robotic dataset in crowded human environments. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 22325–22334

    2024

  55. [55]

    Jongsung Lee, Harin Park, Byeong-Uk Lee, and Kyungdon Joo. 2025. Hush: Holistic panoramic 3d scene understanding using spherical harmonics. InProceedings of the Computer Vision and Pattern Recognition Conference. 16599–16608

    2025

  56. [56]

    Yeonkun Lee, Jaeseok Jeong, Jongseob Yun, Wonjune Cho, and Kuk-Jin Yoon. 2019. Spherephd: Applying cnns on a spherical polyhedron representation of 360deg images. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 9181–9189

    2019

  57. [57]

    Haodong Li, Wangguangdong Zheng, Jing He, Yuhao Liu, Xin Lin, Xin Yang, Ying-Cong Chen, and Chunchao Guo. 2025. DA2: Depth Anything in Any Direction.arXiv preprint arXiv:2509.26618(2025)

    work page arXiv 2025

  58. [58]

    Xiang Li, Haoyuan Cao, Shijie Zhao, Junlin Li, Li Zhang, and Bhiksha Raj. 2023. Panoramic video salient object detection with ambisonic audio guidance. InProceedings of the AAAI Conference on Artificial Intelligence, Vol. 37. 1424–1432

    2023

  59. [59]

    Xuewei Li, Tao Wu, Zhongang Qi, Gaoang Wang, Ying Shan, and Xi Li. 2023. Sgat4pass: Spherical geometry-aware transformer for panoramic semantic segmentation.arXiv preprint arXiv:2306.03403(2023)

    work page arXiv 2023

  60. [60]

    Yuyan Li, Yuliang Guo, Zhixin Yan, Xinyu Huang, Ye Duan, and Liu Ren. 2022. Omnifusion: 360 monocular depth estimation via geometry-aware fusion. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 2801–2810

    2022

  61. [61]

    Zhiqi Li, Wenhai Wang, Hongyang Li, Enze Xie, Chonghao Sima, Tong Lu, Yu Qiao, and Jifeng Dai. 2022. BEVFormer: Learning Bird’s-Eye-View Representation from Multi-Camera Images via Spatiotemporal Transformers.arXiv preprint arXiv:2203.17270(2022)

    work page arXiv 2022

  62. [62]

    Xin Lin, Xian Ge, Dizhe Zhang, Zhaoliang Wan, Xianshun Wang, Xiangtai Li, Wenjie Jiang, Bo Du, Dacheng Tao, Ming-Hsuan Yang, et al. 2025. One flight over the gap: A survey from perspective to panoramic vision.arXiv preprint arXiv:2509.04444(2025)

    work page arXiv 2025

  63. [63]

    Xin Lin, Meixi Song, Dizhe Zhang, Wenxuan Lu, Haodong Li, Bo Du, Ming-Hsuan Yang, Truong Nguyen, and Lu Qi. 2025. Depth Any Panoramas: A Foundation Model for Panoramic Depth Estimation.arXiv preprint arXiv:2512.16913(2025)

    work page arXiv 2025

  64. [64]

    Zekai Lin and Xu Zheng. 2026. PanoEnv: Exploring 3D Spatial Intelligence in Panoramic Environments with Reinforcement Learning.arXiv preprint arXiv:2602.21992(2026)

    work page arXiv 2026

  65. [65]

    Jingguo Liu, Han Yu, Shigang Li, and Jianfeng Li. 2025. 360-degree full-view image segmentation by spherical convolution compatible with large-scale planar pre-trained models. In2025 IEEE International Conference on Multimedia and Expo Workshops (ICMEW). IEEE, 1–6

    2025

  66. [66]

    Kai Luo, Hao Shi, Kunyu Peng, Fei Teng, Sheng Wu, Kaiwei Wang, and Kailun Yang. 2025. OmniTrack++: Omnidirectional Multi-Object Tracking by Learning Large-FoV Trajectory Feedback.arXiv preprint arXiv:2511.00510(2025)

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  67. [67]

    Kai Luo, Hao Shi, Sheng Wu, Fei Teng, Mengfei Duan, Chang Huang, Yuhang Wang, Kaiwei Wang, and Kailun Yang. 2025. Omnidirectional multi-object tracking. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 21959–21969

    2025

  68. [68]

    Chaoxiang Ma, Jiaming Zhang, Kailun Yang, Alina Roitberg, and Rainer Stiefelhagen. 2021. Densepass: Dense panoramic semantic segmentation via unsupervised domain adaptation with attention-augmented context exchange. In2021 IEEE International Intelligent Transportation Systems Conference (ITSC). IEEE, 2766–2772

    2021

  69. [69]

    Koki Maeda, Shuhei Kurita, Taiki Miyanishi, and Naoaki Okazaki. 2023. Query-based image captioning from multi-context 360cdegree images. In Findings of the Association for Computational Linguistics: EMNLP 2023. 6940–6954

    2023

  70. [70]

    Roberto Martin-Martin, Mihir Patel, Hamid Rezatofighi, Abhijeet Shenoi, JunYoung Gwak, Eric Frankel, Amir Sadeghian, and Silvio Savarese. 2021. Jrdb: A dataset and benchmark of egocentric robot visual perception of humans in built environments.IEEE transactions on pattern analysis and machine intelligence45, 6 (2021), 6748–6765

    2021

  71. [71]

    Jeremy Ocampo, Matthew A Price, and Jason D McEwen. 2022. Scalable and equivariant spherical CNNs by discrete-continuous (DISCO) convolutions.arXiv preprint arXiv:2209.13603(2022)

    work page arXiv 2022

  72. [72]

    OpenAI. 2024. GPT-4o System Card. https://cdn.openai.com/gpt-4o-system-card.pdf

    2024

  73. [73]

    Hao Peng, Yun Zhang, and Fang-Lue Zhang. 2025. Robust and enhanced 360 ◦ visual tracking based on dynamic gnomonic projection.Journal of the Royal Society of New Zealand55, 6 (2025), 2169–2197. Manuscript submitted to ACM 34 Zhu and Fan

    2025

  74. [74]

    Nathanaël Perraudin, Michaël Defferrard, Tomasz Kacprzak, and Raphael Sgier. 2019. Deepsphere: Efficient spherical convolutional neural network with healpix sampling for cosmological applications.Astronomy and Computing27 (2019), 130–146

    2019

  75. [75]

    Luigi Piccinelli, Christos Sakaridis, Mattia Segu, Yung-Hsu Yang, Siyuan Li, Wim Abbeloos, and Luc Van Gool. 2025. Unik3d: Universal camera monocular 3d estimation. InProceedings of the Computer Vision and Pattern Recognition Conference. 1028–1039

    2025

  76. [76]

    Giovanni Pintore, Marco Agus, and Enrico Gobbetti. 2020. AtlantaNet: inferring the 3D indoor layout from a single 360 ◦ image beyond the Manhattan world assumption. InEuropean conference on computer vision. Springer, 432–448

    2020

  77. [77]

    Manuel Rey-Area, Mingze Yuan, and Christian Richardt. 2022. 360MonoDepth: High-Resolution 360deg Monocular Depth Estimation. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). 3762–3772

    2022

  78. [78]

    Manolis Savva, Abhishek Kadian, Oleksandr Maksymets, Yili Zhao, Erik Wijmans, Bhavana Jain, Julian Straub, Jia Liu, Vladlen Koltun, Jitendra Malik, et al . 2019. Habitat: A platform for embodied ai research. InProceedings of the IEEE/CVF international conference on computer vision. 9339–9347

    2019

  79. [79]

    Zhijie Shen, Chunyu Lin, Kang Liao, Lang Nie, Zishuo Zheng, and Yao Zhao. 2022. PanoFormer: panorama transformer for indoor 360◦ depth estimation. InEuropean Conference on Computer Vision. Springer, 195–211

    2022

  80. [80]

    Zhijie Shen, Zishuo Zheng, Chunyu Lin, Lang Nie, Kang Liao, Shuai Zheng, and Yao Zhao. 2023. Disentangling orthogonal planes for indoor panoramic room layout estimation with cross-scale distortion awareness. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 17337–17345

    2023

Showing first 80 references.