Research map / Image generation / GANs
2S-ODIS: Two-Stage Omni-Directional Image Synthesis by Geometric Distortion Correction
Atsuya Nakata, Takao Yamanaka
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add the second (high-resolution NFoV refinement) stage (ablation) | 2S-ODIS · SUN360 outdoor test | FID: 28.329 -> 18.263 | Improved |
| add the first (low-resolution global) stage before NFoV synthesis (ablation) | 2S-ODIS · SUN360 outdoor test | FID: 52.453 -> 18.263 | Improved |
| feed decoded low-resolution image (CNN-compressed) to stage two instead of VQGAN codes (ablation) | 2S-ODIS · SUN360 outdoor test | FID: 21.820 -> 18.263 | Improved |
| use pre-trained VQGAN without fine-tuning, with NFoV extraction instead of ERP VQGAN fine-tuning | 2S-ODIS vs OmniDreamer · SUN360 outdoor test | training time: 14 days -> 2-4 days | Improved |
| use parallel MaskGIT-style code prediction instead of auto-regressive prediction | 2S-ODIS vs OmniDreamer · SUN360 | inference time per image: 39.33 seconds -> 1.54 seconds | Improved |
| shorten training from 4 days to 2 days (ablation) | 2S-ODIS · SUN360 outdoor test | FID: 18.263 -> 18.656 | No clear effect |
| compare to OmniDreamer on LPIPS diversity | 2S-ODIS · inpainting conditions | LPIPS: 0.685 vs 0.708 (inpainting) | Hurt |
Later papers that compared with this method
2 later accepted papers report 2S-ODIS as a baseline in their results tables; 2 of them report a better number on most of the results they share with it.
- Panorama Generation From NFoV Image Done Right (CVPR 2025 · better on 2 of 2 shared results)
- Conditional Panoramic Image Generation via Masked Autoregressive Modeling (NeurIPS 2025 · better on 2 of 2 shared results)
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Omni-GAN: On the Secrets of cGANs and Beyond (ICCV 2021 · similarity 0.85)
- Autoregressive Omni-Aware Outpainting for Open-Vocabulary 360-Degree Image Generation (AAAI 2024 · similarity 0.85)
- Fast Omni-Directional Image Super-Resolution: Adapting the Implicit Image Function with Pixel and Semantic-Wise Spherical Geometric Priors (AAAI 2025 · similarity 0.84)
- OmniSplat: Taming Feed-Forward 3D Gaussian Splatting for Omnidirectional Images with Editable Capabilities (CVPR 2025 · similarity 0.83)
- Spherical Pseudo-Cylindrical Representation for Omnidirectional Image Super-resolution (AAAI 2024 · similarity 0.83)
- Omnidirectional Image Super-resolution via Bi-projection Fusion (AAAI 2024 · similarity 0.82)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
