Research map / 3D vision / Depth and stereo
Deep End-to-End Alignment and Refinement for Time-of-Flight RGB-D Module
Di Qiu, Jiahao Pang, Wenxiu Sun, Chengxi Yang
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 flow refinement with camera-parameter depth-to-flow conversion and fusion network (ablation) | cross-modal flow alignment · Synthetic data, sigma 6 | AEPE: 2.70 -> 2.45 | Improved |
| add flow refinement with camera-parameter depth-to-flow conversion and fusion network (ablation) | cross-modal flow alignment · Real data, sigma 6 | AEPE: 1.59 -> 1.36 | Improved |
| feed raw ToF depth directly to the fusion network instead of converting it to flow (ablation) | cross-modal flow alignment · Synthetic data, sigma 6 | AEPE: 2.70 -> 2.79 | Hurt |
| direct depth fusion versus no refinement (ablation) | cross-modal flow alignment · Real data, sigma 2 | AEPE: 1.28 -> 1.29 | No clear effect |
| add gradient loss term (ablation) | ToF-KPN · ToF-FlyingThings3D | MAE (cm): 2.43 -> 1.51 | Improved |
| add kernel normalization (ablation) | ToF-KPN · ToF-FlyingThings3D | MAE (cm): 1.73 -> 1.51 | Improved |
| apply the bias before the kernel instead of after (ablation) | ToF-KPN · ToF-FlyingThings3D | MAE (cm): 1.62 -> 1.51 | Improved |
| add bias term when kernel normalization is absent (ablation) | NoNorm KPN · ToF-FlyingThings3D | MAE (cm): 1.74 -> 1.73 | No clear effect |
| use kernel normalization or early bias over vanilla KPN (ablation) | vanilla KPN (NoNormAftBias) · ToF-FlyingThings3D | MAE (cm): 1.76 -> 1.62 | Improved |
| use kernel prediction (ToF-KPN) instead of direct regression U-Net (ablation) | depth refinement backbone · ToF-FlyingThings3D | MAE (cm): 1.79 -> 1.51 | Improved |
| jointly fine-tune alignment and refinement end to end | DEAR · ToF-FlyingThings3D misaligned test | MAE (cm): 2.90 -> 2.81 | Improved |
| ToF-KPN versus multi-reflection module on FLAT | ToF-KPN · FLAT static test | MAE (cm): 3.88 -> 0.68 | Improved |
Closest papers by meaning
- FloatingFusion: Depth from ToF and Image-Stabilized Stereo Cameras (ECCV 2022 · similarity 0.88)
- Deep End-to-End Time-of-Flight Imaging (CVPR 2018 · similarity 0.87)
- Neural RGB(r)D Sensing: Depth and Uncertainty From a Video Camera (CVPR 2019 · similarity 0.87)
- RGB Guided ToF Imaging System: A Survey of Deep Learning-Based Methods (IJCV 2024 · similarity 0.87)
- RADU: Ray-Aligned Depth Update Convolutions for ToF Data Denoising (CVPR 2022 · similarity 0.86)
- Learnable Burst-Encodable Time-of-Flight Imaging for High-Fidelity Long-Distance Depth Sensing (NeurIPS 2025 · 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.
