Research map / 3D vision / Depth and stereo
NeuralRecon: Real-Time Coherent 3D Scene Reconstruction From Monocular Video
Xi Chen, Jiaming Sun, Yiming Xie, Hujun Bao, Xiaowei Zhou
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 feature-level averaging fusion in place of linear TSDF fusion (ablation) | NeuralRecon · ScanNet (Atlas split) | precision: 0.386 -> 0.432 | Improved |
| add feature-level averaging fusion in place of linear TSDF fusion (ablation) | NeuralRecon · ScanNet (Atlas split) | F-score: 0.462 -> 0.478 | Improved |
| replace average fusion with GRU fusion (ablation) | NeuralRecon · ScanNet (Atlas split) | recall: 0.535 -> 0.572 | Improved |
| replace average fusion with GRU fusion (ablation) | NeuralRecon · ScanNet (Atlas split) | F-score: 0.478 -> 0.488 | Improved |
| fuse within fragment bounding volume instead of occupied area only (ablation) | NeuralRecon with GRU · ScanNet (Atlas split) | recall: 0.572 -> 0.613 | Improved |
| fuse within fragment bounding volume instead of occupied area only (ablation) | NeuralRecon with GRU · ScanNet (Atlas split) | F-score: 0.488 -> 0.494 | Improved |
| increase fragment length from 5 to 9 views (ablation) | NeuralRecon · ScanNet (Atlas split) | F-score: 0.494 -> 0.516 | Improved |
| increase fragment length from 9 to 11 views (ablation) | NeuralRecon · ScanNet (Atlas split) | F-score: 0.516 -> 0.474 | Hurt |
| increase fragment length from 5 to 7 views (ablation) | NeuralRecon · ScanNet (Atlas split) | F-score: 0.494 -> 0.507 | Improved |
| use NeuralRecon volumetric fragment reconstruction instead of Atlas | ScanNet reconstruction · ScanNet (Atlas split, double-layer) | time per key frame (ms): 292 -> 30 | Improved |
| use NeuralRecon instead of Atlas | ScanNet reconstruction · ScanNet (Atlas split, double-layer) | F-score: 0.499 -> 0.516 | Improved |
| use NeuralRecon instead of Atlas | ScanNet reconstruction · ScanNet (Atlas split, double-layer) | recall: 0.732 -> 0.609 | Hurt |
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.
- DPSNet — DPSNet: End-to-end Deep Plane Sweep Stereo (ICLR 2019) · 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
- NeuralRecon: Real-Time Coherent 3D Reconstruction From Monocular Video (CVPR 2021 · similarity 0.99)
- DG-Recon: Depth-Guided Neural 3D Scene Reconstruction (ICCV 2023 · similarity 0.90)
- PlanarRecon: Real-Time 3D Plane Detection and Reconstruction From Posed Monocular Videos (CVPR 2022 · similarity 0.90)
- UniSDF: Unifying Neural Representations for High-Fidelity 3D Reconstruction of Complex Scenes with Reflections (NeurIPS 2024 · similarity 0.88)
- MorpheuS: Neural Dynamic 360deg Surface Reconstruction from Monocular RGB-D Video (CVPR 2024 · similarity 0.87)
- 3D Reconstruction with Generalizable Neural Fields using Scene Priors (ICLR 2024 · 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.
