Research map / 3D vision / Depth and stereo
DoubleTake: Geometry Guided Depth Estimation
Mohamed Sayed, Filippo Aleotti, Jamie Watson, Zawar Qureshi, Guillermo Garcia-Hernando, Gabriel Brostow, Sara Vicente, Michael Firman
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 geometry hint and confidence via Hint MLP to SimpleRecon (ablation) | SimpleRecon · ScanNetV2 depth (incremental) | Abs Diff: .0873 -> .0767 | Improved |
| run offline with two passes using first-pass TSDF as hint | DoubleTake · ScanNetV2 depth | Abs Diff: .0767 -> .0627 | Improved |
| provide no hint at test time (ablation) | DoubleTake · ScanNetV2 depth | Abs Diff: .0873 vs .0870 | No clear effect |
| add confidence map to the Hint MLP input (ablation) | DoubleTake · ScanNetV2 depth (incremental) | Abs Diff: .0863 -> .0767 | Improved |
| feed hint into cost volume encoder instead of Hint MLP (ablation) | DoubleTake · ScanNetV2 depth (incremental) | Abs Diff: .0767 -> .0890 | Hurt |
| use forward-warped previous depth as hint instead of TSDF render (ablation) | DoubleTake · ScanNetV2 depth (incremental) | Abs Diff: .0767 -> .0889 | Hurt |
| replace Hint MLP with cost volume modulation or a single shared MLP (ablation) | DoubleTake · ScanNetV2 depth (incremental) | Abs Diff: .0767 vs .0773 | No clear effect |
| use long-term hints from a previous visit (revisit) | DoubleTake · 3RScan | Abs Diff: .1255 -> .1182 | Improved |
| add pose alignment noise in revisit mode (ablation) | DoubleTake · 3RScan | Abs Diff: .1182 -> .1199 | No clear effect |
| naively input a rendered depth of the global mesh into a depth network | depth estimation network · ScanNetV2 | depth accuracy | No clear effect |
| DoubleTake incremental versus SimpleRecon on 7Scenes delta<1.25 | DoubleTake · 7Scenes | delta < 1.25: 97.33 -> 97.01 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- SimpleRecon — SimpleRecon: 3D Reconstruction without 3D Convolutions (ECCV 2022) · better on 5 of 6 shared results
- FineRecon — FineRecon: Depth-aware Feed-forward Network for Detailed 3D Reconstruction (ICCV 2023) · better on 2 of 3 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
- Neural RGB(r)D Sensing: Depth and Uncertainty From a Video Camera (CVPR 2019 · similarity 0.87)
- Towards Accurate Reconstruction of 3D Scene Shape From A Single Monocular Image (TPAMI 2022 · similarity 0.87)
- A Survey on Deep Learning Techniques for Stereo-Based Depth Estimation (TPAMI 2020 · similarity 0.87)
- Depth2Pose: A Pose-Based Benchmark for Monocular Depth Estimation without Ground-Truth Depth (NeurIPS 2026 · similarity 0.86)
- Depth Pro: Sharp Monocular Metric Depth in Less Than a Second (ICLR 2025 · similarity 0.86)
- A Simple yet Universal Framework for Depth Completion (NeurIPS 2024 · similarity 0.86)
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.
