Research map / 3D vision / Depth and stereo
From Perspective to Fisheye Depth Estimation and Open-Vocabulary Segmentation
Rit Gangopadhyay, Alex Wong
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 Distortion Extenders to frozen backbone | UniDepthV2 · ScanNet++ (real fisheye depth) | RMSE: 0.329 -> 0.200 | Improved |
| add Distortion Extenders to frozen backbone | UniDepthV2 · KITTI-360 (real fisheye depth) | RMSE: 7.093 -> 1.663 | Improved |
| add DEX instead of LoRA | DepthAnything · ScanNet++ (real fisheye depth) | RMSE: 0.706 -> 0.414 | Improved |
| fine-tune with LoRA | DepthAnything · ScanNet++ (real fisheye depth) | RMSE: 0.705 -> 0.706 | No clear effect |
| DEX vs Calibration Tokens | MiDaS · KITTI-360 (real fisheye depth) | delta1: 0.699 -> 0.657 | Hurt |
| DEX vs Calibration Tokens on MiDaS | MiDaS · KITTI-360 (real fisheye depth) | RMSE: 2.588 -> 2.451 | Improved |
| DEX vs LoRA on UniDepthV2 | UniDepthV2 · KITTI-360 (real fisheye depth) | delta1: 0.771 -> 0.842 | Improved |
| apply DEX to UniDepthV2 vs UniK3D | UniDepthV2 · ScanNet++ (real fisheye depth) | RMSE: 0.223 -> 0.200 | Improved |
| apply DEX to open-vocab segmentation | LSeg · WoodScape (open-vocab seg) | mIoU: 0.305 -> 0.362 | Improved |
| apply DEX to CNN-based SED | SED · WoodScape (open-vocab seg) | mIoU: 0.439 -> 0.449 | Improved |
| insert DEX at all blocks vs only the first block (ablation) | DEX (UniDepthV2) · KITTI-360 (real fisheye depth) | RMSE: 1.979 -> 1.663 | Improved |
| use per-layer Extenders instead of one shared set (ablation) | DEX (UniDepthV2) · KITTI-360 (real fisheye depth) | delta1: 0.779 -> 0.842 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- UniK3D — UniK3D: Universal Camera Monocular 3D Estimation (CVPR 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
- Extending Foundational Monocular Depth Estimators to Fisheye Cameras with Calibration Tokens (ICCV 2025 · similarity 0.93)
- Fisheye3R: Adapting Unified 3D Feed-Forward Foundation Models to Fisheye Lenses (ECCV 2026 · similarity 0.91)
- Deep Single Image Camera Calibration by Heatmap Regression to Recover Fisheye Images Under Manhattan World Assumption (CVPR 2024 · similarity 0.85)
- DepthMaster: Unified Monocular Depth Estimation for Perspective and Panoramic Images (NeurIPS 2026 · similarity 0.84)
- Depth2Pose: A Pose-Based Benchmark for Monocular Depth Estimation without Ground-Truth Depth (NeurIPS 2026 · similarity 0.84)
- RSA: Resolving Scale Ambiguities in Monocular Depth Estimators through Language Descriptions (NeurIPS 2024 · similarity 0.84)
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.
