Research map / 3D vision / Depth and stereo
SEAR: Simple and Efficient Adaptation of Visual Geometric Transformers for RGB+Thermal 3D Reconstruction
Vsevolod Skorokhodov, Chenghao Xu, Shuo Sun, Olga Fink, Malcolm Mielle
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 |
|---|---|---|---|
| fine-tune VGGT with LoRA adapters and thermal camera tokens for RGB-T | VGGT · Public datasets (RGB-T) | AUC@30: 22.9 -> 70.0 | Improved |
| fine-tune VGGT with LoRA adapters and thermal camera tokens for RGB-T | VGGT · SEAR dataset (RGB-T) | AUC@30: 23.3 -> 62.8 | Improved |
| fine-tune VGGT with LoRA adapters | VGGT · Public datasets | FPS: 10.46 -> 9.94 | No clear effect |
| use SEAR instead of MINIMA_ROMA | RGB-T pose estimation · Public datasets | FPS: 0.05 -> 9.94 | Improved |
| add learnable thermal camera token (ablation) | SEAR · Public + SEAR datasets | AUC@30: 67.8 -> 68.4 | No clear effect |
| add learnable thermal projector instead of thermal camera token (ablation) | SEAR · Public + SEAR datasets | AUC@30: 68.4 -> 67.2 | Hurt |
| add learnable thermal embedding instead of thermal camera token (ablation) | SEAR · Public + SEAR datasets | AUC@30: 68.4 -> 66.1 | Hurt |
| apply LoRA to frame-attention layers in addition to global-attention (ablation) | SEAR · Public + SEAR datasets | AUC@30: 69.1 -> 68.4 | No clear effect |
| apply LoRA to global-attention layers in addition to frame-attention (ablation) | SEAR · Public + SEAR datasets | AUC@30: 62.4 -> 68.4 | Improved |
| increase LoRA rank r (ablation) | SEAR · Public + SEAR datasets | AUC, RRA, RTA | Improved |
| increase thermal-to-RGB image ratio tau (ablation) | SEAR · camera pose estimation | AUC, RRA, RTA | Mixed |
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.
- VGGT — VGGT: Visual Geometry Grounded Transformer (CVPR 2025) · better on 5 of 6 shared results
- DUSt3R — DUSt3R: Geometric 3D Vision Made Easy (CVPR 2024) · 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
- DarkVGGT: Seeing Through Darkness Using Thermal Geometry without Daylight Tax (NeurIPS 2026 · similarity 0.89)
- TransformerFusion: Monocular RGB Scene Reconstruction using Transformers (NeurIPS 2021 · similarity 0.88)
- Coherent 3D Scene Diffusion From a Single RGB Image (NeurIPS 2024 · similarity 0.87)
- ThermalGen: Style-Disentangled Flow-Based Generative Models for RGB-to-Thermal Image Translation (NeurIPS 2025 · similarity 0.87)
- EAG3R: Event-Augmented 3D Geometry Estimation for Dynamic and Extreme-Lighting Scenes (NeurIPS 2025 · similarity 0.87)
- Fin3R: Fine-tuning Feed-forward 3D Reconstruction Models via Monocular Knowledge Distillation (NeurIPS 2025 · similarity 0.87)
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.
