Salvato in:
| Autori principali: | , , , , , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2024
|
| Soggetti: | |
| Accesso online: | https://arxiv.org/abs/2407.18207 |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866913445990891520 |
|---|---|
| author | Christensen, Anders Mojab, Nooshin Patel, Khushman Ahuja, Karan Akata, Zeynep Winther, Ole Gonzalez-Franco, Mar Colaco, Andrea |
| author_facet | Christensen, Anders Mojab, Nooshin Patel, Khushman Ahuja, Karan Akata, Zeynep Winther, Ole Gonzalez-Franco, Mar Colaco, Andrea |
| contents | Spherical or omni-directional images offer an immersive visual format appealing to a wide range of computer vision applications. However, geometric properties of spherical images pose a major challenge for models and metrics designed for ordinary 2D images. Here, we show that direct application of Fréchet Inception Distance (FID) is insufficient for quantifying geometric fidelity in spherical images. We introduce two quantitative metrics accounting for geometric constraints, namely Omnidirectional FID (OmniFID) and Discontinuity Score (DS). OmniFID is an extension of FID tailored to additionally capture field-of-view requirements of the spherical format by leveraging cubemap projections. DS is a kernel-based seam alignment score of continuity across borders of 2D representations of spherical images. In experiments, OmniFID and DS quantify geometry fidelity issues that are undetected by FID. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2407_18207 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Geometry Fidelity for Spherical Images Christensen, Anders Mojab, Nooshin Patel, Khushman Ahuja, Karan Akata, Zeynep Winther, Ole Gonzalez-Franco, Mar Colaco, Andrea Computer Vision and Pattern Recognition Machine Learning Spherical or omni-directional images offer an immersive visual format appealing to a wide range of computer vision applications. However, geometric properties of spherical images pose a major challenge for models and metrics designed for ordinary 2D images. Here, we show that direct application of Fréchet Inception Distance (FID) is insufficient for quantifying geometric fidelity in spherical images. We introduce two quantitative metrics accounting for geometric constraints, namely Omnidirectional FID (OmniFID) and Discontinuity Score (DS). OmniFID is an extension of FID tailored to additionally capture field-of-view requirements of the spherical format by leveraging cubemap projections. DS is a kernel-based seam alignment score of continuity across borders of 2D representations of spherical images. In experiments, OmniFID and DS quantify geometry fidelity issues that are undetected by FID. |
| title | Geometry Fidelity for Spherical Images |
| topic | Computer Vision and Pattern Recognition Machine Learning |
| url | https://arxiv.org/abs/2407.18207 |