Research map / Architectures and efficiency / CNN design
Long Range Pooling for 3D Large-Scale Scene Understanding
Read the paperCheck 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 |
|---|---|---|---|
| Add the LRP module after each stage | VMNet U-Net · ScanNet v2 val | mIoU (%): 71.2 -> 75.0 | Helped |
| Add max pooling without dilation or receptive-field selection (ablation) | VMNet U-Net · ScanNet v2 val | mIoU (%): 71.2 -> 72.6 | Helped |
| Add dilation to max pooling without receptive-field selection (ablation) | VMNet U-Net · ScanNet v2 val | mIoU (%): 72.6 -> 73.7 | Helped |
| Add receptive-field selection to max pooling without dilation (ablation) | VMNet U-Net · ScanNet v2 val | mIoU (%): 72.6 -> 73.1 | Helped |
| Use both dilated max pooling and receptive-field selection (ablation) | VMNet U-Net · ScanNet v2 val | mIoU (%): 73.7 -> 75.0 | Helped |
| Place LRP before stages rather than after them (ablation) | VMNet U-Net · ScanNet v2 val | mIoU (%): 75.0 -> 74.0 | Hurt |
| Place LRP in the middle of stages rather than after them (ablation) | VMNet U-Net · ScanNet v2 val | mIoU (%): 75.0 -> 73.4 | Hurt |
| Place LRP in a parallel branch rather than after stages (ablation) | VMNet U-Net · ScanNet v2 val | mIoU (%): 75.0 -> 73.1 | Hurt |
| Expand the max-pooling receptive-field range from ×3 to ×27 (ablation) | LRP · ScanNet v2 val | mIoU (%): 72.7 -> 73.9 | Helped |
| Use max pooling at ×3, ×9 and ×27 instead of ×27 only (ablation) | LRP · ScanNet v2 val | mIoU (%): 73.9 -> 75.0 | Helped |
| Replace dilation max pooling with dilation average pooling across the tested ranges (ablation) | LRP · ScanNet v2 val | mIoU (%): 75.0 -> 73.9 | Hurt |
| Replace dilation max pooling with dilation convolution across the tested ranges (ablation) | LRP · ScanNet v2 val | mIoU (%): 75.0 -> 73.8 | Hurt |
Closest papers by meaning
- LSK3DNet: Towards Effective and Efficient 3D Perception with Large Sparse Kernels (CVPR 2024 · similarity 0.85)
- A General Protocol to Probe Large Vision Models for 3D Physical Understanding (NeurIPS 2024 · similarity 0.85)
- Large Spatial Model: End-to-end Unposed Images to Semantic 3D (NeurIPS 2024 · similarity 0.84)
- CAM Back Again: Large Kernel CNNs from a Weakly Supervised Object Localization Perspective (CVPR 2024 · similarity 0.84)
- Beyond Skip Connection: Pooling and Unpooling Design for Elimination Singularities (AAAI 2025 · similarity 0.84)
- DL3DV-10K: A Large-Scale Scene Dataset for Deep Learning-based 3D Vision (CVPR 2024 · 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.