Research map / 3D vision / Depth and stereo
Dark3R: Learning Structure from Motion in the Dark
Andrew Y. Guo, Anagh Malik, SaiKiran Tedla, Yutong Dai, Yiqian Qin, Zach Salehe, Benjamin Attal, Sotiris Nousias, Kiriakos N. Kutulakos, David B. Lindell
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 |
|---|---|---|---|
| use LoRA instead of full fine-tuning (ablation) | Dark3R · low-light SfM (120 images) | ATE: 0.476 -> 0.050 | Improved |
| use raw inputs instead of sRGB (ablation) | Dark3R · low-light SfM (120 images) | ATE: 0.087 -> 0.050 | Improved |
| supervise with clean raw pairs instead of clean sRGB pairs (ablation) | Dark3R · low-light SfM (120 images) | ATE: 0.064 -> 0.050 | Mixed |
| train on both simulated and captured data instead of simulated only (ablation) | Dark3R · low-light SfM (120 images) | ATE: 0.078 -> 0.050 | Improved |
| add L_clean regularization on clean pairs (ablation) | Dark3R · low-light SfM (120 images) | ATE: 0.051 -> 0.050 | No clear effect |
| fine-tune encoder, decoder and head vs encoder only (ablation) | Dark3R · low-light SfM (120 images) | ATE / RPE R: ATE 0.030 -> 0.050; RPE R 0.149 -> 0.121 | Mixed |
| Dark3R teacher-student adaptation vs MASt3R-SfM | MASt3R-SfM · low-light SfM (120 images) | ATE: 0.088 -> 0.050 | Improved |
| scale from 120 to full image set (bundle adjustment relied on more) | Dark3R · low-light SfM | ATE: 0.050 -> 0.139 | Hurt |
| add depth supervision from Dark3R pointmaps (ablation) | Dark3R-NeRF · low-light view synthesis | PSNR: 34.91 -> 36.17 | Improved |
| avoid black level clipping (ablation) | Dark3R-NeRF · low-light view synthesis | PSNR: 34.98 -> 36.17 | Improved |
| add stochastic preconditioning (ablation) | Dark3R-NeRF · low-light view synthesis | PSNR: 36.05 -> 36.17 | Improved |
| Dark3R-NeRF vs LE3D on SSIM | Dark3R-NeRF · low-light view synthesis | SSIM: 0.878 -> 0.866 | Hurt |
Later papers that compared with this method
1 later accepted paper reports Dark3R as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- 3R-Adapter: Retrieval, Rewiring, and Refinement for Efficient Adaptation of 3D Reconstruction Model (NeurIPS 2026 · better on 1 of 1 shared result)
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.
- VGGT — VGGT: Visual Geometry Grounded Transformer (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
- Light3R-SfM: Towards Feed-forward Structure-from-Motion (CVPR 2025 · similarity 0.86)
- Lighting Every Darkness with 3DGS: Fast Training and Real-Time Rendering for HDR View Synthesis (NeurIPS 2024 · similarity 0.86)
- 4D3R: Motion-Aware Neural Reconstruction and Rendering of Dynamic Scenes from Monocular Videos (NeurIPS 2025 · similarity 0.86)
- DarkVGGT: Seeing Through Darkness Using Thermal Geometry without Daylight Tax (NeurIPS 2026 · similarity 0.86)
- Sparse Views Near Light: A Practical Paradigm for Uncalibrated Point-light Photometric Stereo (CVPR 2024 · similarity 0.85)
- SPARS3R: Semantic Prior Alignment and Regularization for Sparse 3D Reconstruction (CVPR 2025 · similarity 0.85)
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.
