Research map / Detection and segmentation / Open-set and few-shot detection
3D Compositional Zero-Shot Learning with DeCompositional Consensus
Muhammad Ferjad Naeem, Evin Pınar Örnek, Yongqin Xian, Luc Van Gool, Federico Tombari
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 DCC consensus scoring instead of direct segmentation | PointNet · C-PartNet segmentation | mIoU harmonic mean: 28.5 -> 35.2 | Improved |
| add DCC consensus scoring instead of direct segmentation | PointNet · C-PartNet segmentation (unseen) | mIoU: 20.1 -> 32.7 | Improved |
| add DCC instead of direct segmentation | PointNet · C-PartNet classification | accuracy harmonic mean: 55.9 vs 33.1 for CGE | Improved |
| add L_DeComp loss to L_Part loss (ablation) | DCC · C-PartNet classification | accuracy harmonic mean: 14.9 -> 72.8 | Improved |
| add L_Part loss to L_DeComp loss (ablation) | DCC · C-PartNet classification | accuracy harmonic mean: 29.0 -> 72.8 | Improved |
| learn consensus on predicted segmentation instead of replacing with ground-truth mask in hypothesis bank (ablation) | DCC · C-PartNet classification | accuracy harmonic mean: 52.6 -> 72.8 | Improved |
| learn consensus on predicted segmentation instead of ground-truth mask (ablation) | DCC · C-PartNet segmentation | mIoU harmonic mean: 39.1 -> 45.1 | Improved |
| add DCC to DGCNN (ablation) | DGCNN · C-PartNet segmentation | mIoU harmonic mean: 29.5 -> 36.2 | Improved |
| add DCC to ConvPoint (ablation) | ConvPoint · C-PartNet segmentation | mIoU harmonic mean: 2.9 -> 29.5 | Improved |
| add DCC to GDANet (ablation) | GDANet · C-PartNet segmentation | mIoU harmonic mean: 28.7 -> 33.5 | Improved |
| compute L_Seg over object part prior instead of all parts (ablation) | PointNet oracle · C-PartNet oracle segmentation | mIoU harmonic mean: 43.2 -> 47.9 | Improved |
| compute L_Seg over object part prior instead of all parts (ablation) | ConvPoint oracle · C-PartNet oracle segmentation | mIoU harmonic mean: 29.1 -> 43.5 | Improved |
Closest papers by meaning
- PartSLIP: Low-Shot Part Segmentation for 3D Point Clouds via Pretrained Image-Language Models (CVPR 2023 · similarity 0.89)
- Zero-Shot Point Cloud Segmentation by Semantic-Visual Aware Synthesis (ICCV 2023 · similarity 0.88)
- See More and Know More: Zero-shot Point Cloud Segmentation via Multi-modal Visual Data (ICCV 2023 · similarity 0.88)
- PartNet: A Large-Scale Benchmark for Fine-Grained and Hierarchical Part-Level 3D Object Understanding (CVPR 2019 · similarity 0.88)
- 3DCoMPaT200: Language Grounded Large-Scale 3D Vision Dataset for Compositional Recognition (NeurIPS 2024 · similarity 0.87)
- PartNeXt: A Next-Generation Dataset for Fine-Grained and Hierarchical 3D Part Understanding (NeurIPS 2025 · similarity 0.86)
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.
