Research map / Point clouds / Segmentation and detection
Point Cloud Scene Completion With Joint Color and Semantic Estimation From Single RGB-D Image
Zhaoxuan Zhang, Xiaoguang Han, Bo Dong, Tong Li, Baocai Yin, Xin She 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 RGB input and color prediction (CSC vs GSC) | DQN scene completion · 3D-FUTURE test | Chamfer Distance: 0.5364 -> 0.4503 | Improved |
| add semantic segmentation output (CSSC vs CSC) | DQN scene completion · 3D-FUTURE test | Chamfer Distance: 0.4503 -> 0.4200 | Improved |
| replace DQN with A3C for view path planning (ablation) | CSSC · 3D-FUTURE test | Chamfer Distance: 0.4200 -> 0.4095 | Improved |
| A3C selects fewer viewpoints than DQN (ablation) | CSSC · 3D-FUTURE test | average viewpoints: 7.83 -> 5.31 | Improved |
| use RL view planning instead of uniform sampling of 5 views (ablation) | Ours · 3D-FUTURE test | Chamfer Distance: 0.4520 -> 0.4095 | Improved |
| use RL view planning instead of uniform 20 views (ablation) | Ours · 3D-FUTURE test | Chamfer Distance: 0.4576 -> 0.4095 | Improved |
| add hole reward R_hole to A3C (ablation) | Ours · 3D-FUTURE test | Chamfer Distance: 0.4508 -> 0.4095 | Improved |
| add point-cloud accuracy reward R_pcacc (ablation) | Ours · 3D-FUTURE test | Chamfer Distance: 0.4423 -> 0.4095 | Improved |
| shrink phi range of the action space to +-25 deg (ablation) | Ours · 3D-FUTURE test | Chamfer Distance: 0.4095 -> 0.5231 | Hurt |
| change theta from 60/80 to 70/90 deg (ablation) | Ours · 3D-FUTURE test | Chamfer Distance: 0.5231 -> 0.5187 | No clear effect |
| add segmentation input to depth inpainting (PartConv) (ablation) | depth inpainting · 3D-FUTURE depth inpainting | L1 loss: 0.0805 -> 0.0653 | Improved |
| add projection back-propagation to depth inpainting (ablation) | depth inpainting · 3D-FUTURE depth inpainting | L1 loss: 0.0593 -> 0.0543 | Improved |
Closest papers by meaning
- Deep Reinforcement Learning of Volume-Guided Progressive View Inpainting for 3D Point Scene Completion From a Single Depth Image (CVPR 2019 · similarity 0.93)
- Point Cloud Semantic Scene Completion from RGB-D Images (AAAI 2021 · similarity 0.90)
- Rethinking Multimodal Point Cloud Completion: A Completion-by-Correction Perspective (AAAI 2026 · similarity 0.90)
- RGBD Based Dimensional Decomposition Residual Network for 3D Semantic Scene Completion (CVPR 2019 · similarity 0.88)
- Deep Depth Completion of a Single RGB-D Image (CVPR 2018 · similarity 0.88)
- GenPC: Zero-shot Point Cloud Completion via 3D Generative Priors (CVPR 2025 · similarity 0.86)
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.
