Research map / Point clouds / Segmentation and detection
SLiDE: Self-Supervised LiDAR De-Snowing through Reconstruction Difficulty
Gwangtak Bae, Byungjun Kim, Seongyong Ahn, Jihong Min, Inwook Shim
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 multi-hypothesis learning (3 hypotheses) to PR-Net (ablation) | Ours · Synthetic snow noise data (nuScenes base) | IoU: 65.75 -> 79.62 | Improved |
| use 3 hypotheses instead of 2 (ablation) | Ours · Synthetic snow noise data (nuScenes base) | IoU: 76.17 -> 79.62 | Improved |
| use 4 hypotheses instead of 3 (ablation) | Ours · Synthetic snow noise data (nuScenes base) | IoU: 79.62 -> 76.52 | Hurt |
| blank 50% of points instead of 10% (ablation) | Ours · Synthetic snow noise data (nuScenes base) | IoU: 76.64 -> 79.62 | Improved |
| blank 70% instead of 50% (ablation) | Ours · Synthetic snow noise data (nuScenes base) | IoU: 79.62 -> 76.57 | Hurt |
| use 20th percentile shift instead of minimum output (ablation) | Ours · Synthetic snow noise data (nuScenes base) | IoU: 72.63 -> 79.62 | Improved |
| use 20th percentile shift instead of 10th (ablation) | Ours · Synthetic snow noise data (nuScenes base) | IoU: 79.36 -> 79.62 | No clear effect |
| add semi-supervised extension with 239 labels vs WeatherNet | Ours Semi-sup · Synthetic snow noise data (nuScenes base) | IoU: 41.40 -> 82.44 | Improved |
| add semi-supervised extension at full labels | Ours Semi-sup · Synthetic snow noise data (nuScenes base) | IoU: 84.04 -> 84.24 | No clear effect |
| use pretrain weighting instead of smooth transfer at 1% labels (ablation) | Semi-supervised · Synthetic snow noise data (nuScenes base) (1% labels) | IoU: 82.44 -> 62.46 | Hurt |
| use ramp up/down weighting at 100% labels instead of supervised only (ablation) | Semi-supervised · Synthetic snow noise data (nuScenes base) (100% labels) | IoU: 84.04 -> 82.57 | Hurt |
| use label-free method vs DROR | Ours · Synthetic snow noise data (nuScenes base) | IoU: 33.68 -> 79.62 | Improved |
Closest papers by meaning
- Complete & Label: A Domain Adaptation Approach to Semantic Segmentation of LiDAR Point Clouds (CVPR 2021 · similarity 0.87)
- Implicit Surface Contrastive Clustering for LiDAR Point Clouds (CVPR 2023 · similarity 0.87)
- Less Is More: Reducing Task and Model Complexity for 3D Point Cloud Semantic Segmentation (CVPR 2023 · similarity 0.86)
- LESS: Label-Efficient Semantic Segmentation for LiDAR Point Clouds (ECCV 2022 · similarity 0.86)
- Learning Graph-Convolutional Representations for Point Cloud Denoising (ECCV 2020 · similarity 0.86)
- Learnable Earth Parser: Discovering 3D Prototypes in Aerial Scans (CVPR 2024 · similarity 0.85)
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.
