Research map / Point clouds / Shape and surface completion
RL-GAN-Net: A Reinforcement Learning Agent Controlled GAN Network for Real-Time Point Cloud Shape Completion
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| complete partial inputs with RL-GAN-Net before PointNet classification | PointNet · ShapeNetCore, 70% missing | classification accuracy: 50.2 -> 83.8 | Helped |
| add the AE/discriminator switch to vanilla RL-GAN-Net (ablation) | RL-GAN-Net · ShapeNetCore, 20% missing | classification accuracy: 97.7 -> 98.1 | Hurt |
| add the AE/discriminator switch to vanilla RL-GAN-Net (ablation) | RL-GAN-Net · ShapeNetCore, 30%, 40%, 50%, and 70% missing | classification accuracy: 96.7 -> 97.2; 95.0 -> 95.5; 92.7 -> 93.3; 82.5 -> 83.8 | Helped |
| add the AE/discriminator switch to vanilla RL-GAN-Net (ablation) | RL-GAN-Net · Shape completion, missing data | Chamfer distance | Helped |
| use GFV loss alone instead of combining reward losses (ablation) | RL-GAN-Net · ShapeNetCore shape completion | semantic category completion | Helped |
| use Chamfer loss alone instead of combining reward losses (ablation) | RL-GAN-Net · ShapeNetCore shape completion | shape completion quality | Hurt |
| use discriminator loss alone instead of combining reward losses (ablation) | RL-GAN-Net · ShapeNetCore shape completion | input-consistent semantic completion | Hurt |
| combine Chamfer loss with GFV loss (ablation) | RL-GAN-Net · ShapeNetCore shape completion | shape alignment | Helped |
| use a 6- or 32-dimensional GAN z-vector instead of a 1-dimensional z-vector (ablation) | RL-GAN-Net · ShapeNetCore shape completion | completion performance | No clear effect |
| replace per-input optimization with RL control of the GAN seed | latent-GAN shape completion · Shape completion inference | runtime: 324 seconds -> approximately one millisecond | Helped |
Closest papers by meaning
- Learning Representations and Generative Models for 3D Point Clouds (ICML 2018 · similarity 0.86)
- CPCGAN: A Controllable 3D Point Cloud Generative Adversarial Network with Semantic Label Generating (AAAI 2021 · similarity 0.86)
- Diverse Shape Completion via Style Modulated Generative Adversarial Networks (NeurIPS 2023 · similarity 0.86)
- CDPNet: Cross-Modal Dual Phases Network for Point Cloud Completion (AAAI 2024 · similarity 0.84)
- DC-PCN: Point Cloud Completion Network with Dual-Codebook Guided Quantization (AAAI 2025 · similarity 0.83)
- What You See is What You GAN: Rendering Every Pixel for High-Fidelity Geometry in 3D GANs (CVPR 2024 · similarity 0.83)
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.