Research map / Detection and segmentation / 3D detection and driving
Generative Sparse Detection Networks for 3D Single-shot Object Detection
JunYoung Gwak, Christopher Choy, Silvio Savarese
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 balanced cross entropy to the sparsity loss (ablation) | GSDN with CE losses · ScanNet v2 val (60k iter) | mAP@0.25: 33.1 -> 50.7 | Improved |
| add balanced cross entropy to the anchor loss (ablation) | GSDN with BCE sparsity · ScanNet v2 val (60k iter) | mAP@0.25: 50.7 -> 57.2 | Improved |
| add balanced cross entropy to the sparsity loss (ablation) | GSDN with CE losses · ScanNet v2 val (60k iter) | mAP@0.5: 9.26 -> 25.4 | Improved |
| add balanced cross entropy to the anchor loss (ablation) | GSDN with BCE sparsity · ScanNet v2 val (60k iter) | mAP@0.5: 25.4 -> 29.7 | Improved |
| deepen encoder from ResNet14 to ResNet34 (ablation) | GSDN · ScanNet v2 val (60k iter) | mAP@0.25: 52.1 -> 57.2 | Improved |
| deepen encoder from ResNet14 to ResNet34 (ablation) | GSDN · ScanNet v2 val (60k iter) | mAP@0.5: 25.4 -> 29.7 | Improved |
| deepen encoder from ResNet18 to ResNet34 (ablation) | GSDN · ScanNet v2 val (60k iter) | mAP@0.25: 55.1 -> 57.2 | Improved |
| use five anchor ratios instead of one (ablation) | GSDN · ScanNet v2 val (60k iter) | mAP@0.5: 22.7 -> 29.7 | Improved |
| use five anchor ratios instead of one (ablation) | GSDN · ScanNet v2 val (60k iter) | mAP@0.25: 56.3 -> 57.2 | Improved |
| use five anchor ratios instead of three (ablation) | GSDN · ScanNet v2 val (60k iter) | mAP@0.25: 55.3 -> 57.2 | Improved |
| lower sparsity pruning confidence tau (ablation) | GSDN · ScanNet v2 val | mAP@0.25 | Improved |
| replace two-stage VoteNet with single-shot GSDN | VoteNet · ScanNet v2 val | mAP@0.25: 58.6 -> 62.8 | Improved |
Closest papers by meaning
- Spatial Pruned Sparse Convolution for Efficient 3D Object Detection (NeurIPS 2022 · similarity 0.90)
- Focal Sparse Convolutional Networks for 3D Object Detection (CVPR 2022 · similarity 0.90)
- FSHNet: Fully Sparse Hybrid Network for 3D Object Detection (CVPR 2025 · similarity 0.89)
- SparseGNV: Generating Novel Views of Indoor Scenes with Sparse RGB-D Images (AAAI 2024 · similarity 0.88)
- SAFDNet: A Simple and Effective Network for Fully Sparse 3D Object Detection (CVPR 2024 · similarity 0.87)
- OA-CNNs: Omni-Adaptive Sparse CNNs for 3D Semantic Segmentation (CVPR 2024 · 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.
