Research map / 3D vision / Face and shape
TriNLOS: Triplane Representations for Neural Non-Line-of-Sight Imaging
Bonmu Do, Ji Hyun Nam
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 Restormer attention (ablation) | TriNLOS · Synthetic Seen | PSNR: 28.44 -> 30.20 | Improved |
| add axis cross-attention (ablation) | TriNLOS · Synthetic Seen | PSNR: 29.21 -> 30.20 | Improved |
| add ResConvNeXt blocks (ablation) | TriNLOS · Synthetic Seen | PSNR: 29.31 -> 30.20 | Improved |
| use ELCT instead of FK initialization (ablation) | TriNLOS · Synthetic Seen | PSNR: 29.25 -> 30.20 | Improved |
| use ELCT instead of LCT initialization (ablation) | TriNLOS · Synthetic Seen | PSNR: 29.20 -> 30.20 | Improved |
| use ELCT instead of Phasor initialization (ablation) | TriNLOS · Synthetic Seen | PSNR: 28.95 -> 30.20 | Improved |
| block depth gradients from the backbone (ablation) | TriNLOS · Synthetic Seen | PSNR: 28.47 -> 30.20 | Improved |
| keep yz projection (ablation) | TriNLOS · Synthetic Seen | PSNR: 28.98 -> 30.26 | Improved |
| keep xz projection (ablation) | TriNLOS · Synthetic Seen | PSNR: 29.30 -> 30.26 | Improved |
| keep xy projection (ablation) | TriNLOS · Synthetic Seen | PSNR: 29.44 -> 30.26 | Improved |
| triplane backbone vs NLOST | TriNLOS · Synthetic Unseen | PSNR: 25.65 -> 26.13 | Improved |
| triplane backbone vs NLOST | TriNLOS · Synthetic Seen | runtime (ms): 165.29 -> 103.22 | 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.
- NLOST — NLOST: Non-Line-of-Sight Imaging With Transformer (CVPR 2023) · better on 5 of 5 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
- Enhancing Non-line-of-sight Imaging via Learnable Inverse Kernel and Attention Mechanisms (ICCV 2023 · similarity 0.89)
- Physics to the Rescue: Deep Non-Line-of-Sight Reconstruction for High-Speed Imaging (TPAMI 2022 · similarity 0.88)
- Dual-branch Graph Feature Learning for NLOS Imaging (AAAI 2025 · similarity 0.86)
- Virtual Scanning: Unsupervised Non-line-of-sight Imaging from Irregularly Undersampled Transients (NeurIPS 2024 · similarity 0.85)
- Real-time Acquisition and Reconstruction of Dynamic Volumes with Neural Structured Illumination (CVPR 2024 · similarity 0.84)
- PlatoNeRF: 3D Reconstruction in Plato's Cave via Single-View Two-Bounce Lidar (CVPR 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.
