Research map / Point clouds / Point cloud classification
3DCNN-DQN-RNN: A Deep Reinforcement Learning Framework for Semantic Parsing of Large-Scale 3D Point Clouds
Fangyu Liu, Shuaipeng Li, Liqiang Zhang, Chenghu Zhou, Rongtian Ye, Yuebin Wang, Jiwen Lu
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 Network 2 (Residual RNN) after Network 1 (ablation) | 3DCNN-DQN · Stanford 3D semantic parsing | mean class accuracy (%): 57.28 -> 70.76 | Improved |
| add Network 1 (3DCNN-DQN eye window) before Network 2 (ablation) | Residual RNN · Stanford 3D semantic parsing | mean class accuracy (%): 36.63 -> 70.76 | Improved |
| add Network 2 to fix wall class (ablation) | 3DCNN-DQN · Stanford 3D, wall class | accuracy (%): 24.10 -> 60.08 | Improved |
| add Network 2 for door class (ablation) | 3DCNN-DQN · Stanford 3D, door class | accuracy (%): 91.48 -> 33.41 | Hurt |
| add Network 2 for table class (ablation) | 3DCNN-DQN · Stanford 3D, table class | accuracy (%): 77.41 -> 70.48 | No clear effect |
| use proposed N1+N2 instead of Armeni et al. (S1) | 3DCNN-DQN-RNN · Stanford 3D semantic parsing | mean class accuracy (%): 49.93 -> 70.76 | Improved |
| use proposed N1+N2 instead of 3D FCN-1 | 3DCNN-DQN-RNN · Stanford 3D semantic parsing | mean class accuracy (%): 62.96 -> 70.76 | Improved |
| classify twelve classes at once with 3D FCN rather than one class (ablation) | 3D FCN (VGG-16) · Stanford 3D semantic parsing | mean class accuracy (%): 62.96 -> 12.97 | Hurt |
| add winner replay to the DQN (ablation) | 3DCNN-DQN · Stanford 3D, search time | searching time reduction: reduced by 73% | Improved |
| run the DQN longer (ablation) | 3DCNN-DQN · Stanford 3D semantic parsing | classification accuracy | Improved |
Closest papers by meaning
- Fully-Convolutional Point Networks for Large-Scale Point Clouds (ECCV 2018 · similarity 0.90)
- PointNet: Deep Learning on Point Sets for 3D Classification and Segmentation (CVPR 2017 · similarity 0.89)
- 3D Recurrent Neural Networks with Context Fusion for Point Cloud Semantic Segmentation (ECCV 2018 · similarity 0.89)
- Learning of 3D Graph Convolution Networks for Point Cloud Analysis (TPAMI 2021 · similarity 0.88)
- DeepLA-Net: Very Deep Local Aggregation Networks for Point Cloud Analysis (CVPR 2025 · similarity 0.87)
- Exploring Vision Semantic Prompt for Efficient Point Cloud Understanding (ICML 2025 · 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.
