Research map / Image generation / Autoregressive and latent generation
Tangent Images for Mitigating Spherical Distortion
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| represent spherical inputs as tangent images and use standard 2D convolutions | ModelNet40 shape classification · ModelNet40 | accuracy: 89.1% vs 90.5% | Mixed |
| use level 10 tangent-image input rather than level 5 | Stanford2D3DS semantic segmentation, RGB-D · Stanford2D3DS, RGB-D | mAcc: 50.9 -> 70.1 | Helped |
| use level 10 tangent-image input rather than level 5 | Stanford2D3DS semantic segmentation, RGB-D · Stanford2D3DS, RGB-D | mIOU: 38.3 -> 52.5 | Helped |
| use base level 1 instead of base level 0 (ablation) | Level 10 RGB semantic segmentation · Stanford2D3DS | mAcc: 61.0 -> 65.2 | Helped |
| use base level 1 instead of base level 0 (ablation) | Level 10 RGB semantic segmentation · Stanford2D3DS | mIOU: 44.3 -> 45.6 | Helped |
| use base level 2 instead of base level 1 (ablation) | Level 10 RGB semantic segmentation · Stanford2D3DS | mIOU: 45.6 -> 42.7 | Hurt |
| fine-tune the transferred model on spherical data for 1 epoch (ablation) | Perspective-trained FCN-ResNet101 · Stanford2D3DS transfer, level 8 | mIOU: 47.3 -> 51.9 | Helped |
| fine-tune the transferred model on spherical data for 10 epochs (ablation) | Perspective-trained FCN-ResNet101 · Stanford2D3DS transfer, level 8 | mIOU: 47.3 -> 54.6 | Helped |
| fine-tune on spherical data instead of evaluating without fine-tuning (ablation) | Perspective-trained network · OmniSYNTHIA, level 8 | mAcc: 70.8 -> 73.2 | Helped |
| use tangent images rather than equirectangular images for feature detection | SIFT matching, Hard split · Spherical keypoints, Hard | PMR: 22.2% -> 30.1% | Helped |
| use tangent images rather than equirectangular images for feature detection | SIFT matching, Easy split · Spherical keypoints, Easy | MS: 13.6% -> 17.7% | Helped |
| increase keypoint tangent-image base level beyond level 1 (ablation) | SIFT matching · Spherical keypoints, Hard and Easy | PMR: 30.1% -> 27.4%; 34.6% -> 31.9% | Hurt |
Closest papers by meaning
- Learning Spherical Convolution for Fast Features from 360° Imagery (NeurIPS 2017 · similarity 0.88)
- Unified Spherical Frontend: Learning Rotation-Equivariant Representations of Spherical Images from Any Camera (CVPR 2026 · similarity 0.87)
- BRISKS: Binary Features for Spherical Images on a Geodesic Grid (CVPR 2017 · similarity 0.87)
- Rotation Equivariant Graph Convolutional Network for Spherical Image Classification (CVPR 2020 · similarity 0.87)
- Learning Shape-Independent Transformation via Spherical Representations for Category-Level Object Pose Estimation (ICLR 2025 · similarity 0.84)
- SphereDiffusion: Spherical Geometry-Aware Distortion Resilient Diffusion Model (AAAI 2024 · similarity 0.83)
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.