Research map / Architectures and efficiency / Attention and transformers
3D Reconstruction from Transient Measurements with Time-Resolved Transformer
Yue Li, Shida Sun, Hong Wu Yu, Feng Ying Xu, Zhiwei Xiong
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 TV regularization loss (ablation) | TRT-LOS · Simulated LOS (SBR X:2 avg) | depth RMSE: 0.0370 -> 0.0361 | Improved |
| add TV regularization loss (ablation) | TRT-LOS · Simulated LOS (SBR Y:100 avg) | depth RMSE: 0.0876 -> 0.0813 | Improved |
| add deep local features to global-only fusion (ablation) | TRT-LOS · Simulated LOS (SBR X:2 avg) | depth RMSE: 4.7183 -> 0.0361 | Improved |
| add deep global features to local-only fusion (ablation) | TRT-LOS · Simulated LOS (SBR X:2 avg) | depth RMSE: 0.0389 -> 0.0361 | Improved |
| add temporal attention to spatial-only attention (ablation) | TRT-LOS · Simulated LOS (SBR X:2 avg) | depth RMSE: 0.0994 -> 0.0361 | Improved |
| add spatial attention to temporal-only attention (ablation) | TRT-LOS · Simulated LOS (SBR X:2 avg) | depth RMSE: 0.0392 -> 0.0361 | Improved |
| add local-global integration (LGInt) in cross attention decoder (ablation) | TRT-LOS · Simulated LOS (SBR Y:100 avg) | depth RMSE: 0.0892 -> 0.0813 | Improved |
| increase TRT blocks from 12 to 24 (ablation) | TRT-LOS · Simulated LOS (SBR X:50 avg) | depth RMSE: 0.0508 -> 0.0540 | Hurt |
| increase TRT blocks from 12 to 24 (ablation) | TRT-LOS · Simulated LOS (SBR X:2 avg) | depth RMSE: 0.0361 -> 0.0360 | No clear effect |
| increase TRT input channels from 64 to 128 (ablation) | TRT-LOS · Simulated LOS (SBR X:2 avg) | depth RMSE: 0.0361 -> 0.0374 | Hurt |
| add local-global integration (LGInt) in cross attention decoder (ablation) | TRT-NLOS · Simulated NLOS Seen | depth RMSE: 0.0904 -> 0.0849 | Improved |
| add temporal attention to spatial-only attention (ablation) | TRT-NLOS · Simulated NLOS Seen | depth RMSE: 0.0877 -> 0.0849 | Improved |
Closest papers by meaning
- NLOST: Non-Line-of-Sight Imaging With Transformer (CVPR 2023 · similarity 0.90)
- Deep Non-Line-of-Sight Reconstruction (CVPR 2020 · similarity 0.88)
- Separating the Wheat from the Chaff: Spatio-Temporal Transformer with View-interweaved Attention for Photon-Efficient Depth Sensing (AAAI 2025 · similarity 0.87)
- Virtual Scanning: Unsupervised Non-line-of-sight Imaging from Irregularly Undersampled Transients (NeurIPS 2024 · similarity 0.87)
- Toward Dynamic Non-Line-of-Sight Imaging with Mamba Enforced Temporal Consistency (NeurIPS 2024 · similarity 0.86)
- Towards 3D Vision with Low-Cost Single-Photon Cameras (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.
