Research map / Detection and segmentation / 3D detection and driving
Objects Can Move: 3D Change Detection by Geometric Transformation Consistency
Aikaterini Adam, Torsten Sattler, Konstantinos Karantzalos, Tomas Pajdla
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 geometric-transformation graph cut optimization after initial depth-based detection (ablation) | Palazzolo et al. initial detection · 3RScan change detection | mean IoU (%): 54.23% -> 68.40% | Improved |
| add geometric-transformation graph cut optimization after initial depth-based detection (ablation) | Palazzolo et al. initial detection · 3RScan change detection | mean recall (%): 31.48% -> 76.05% | Improved |
| propagate change by geometric consistency instead of photoconsistency | Taneja et al. · 3RScan change detection | mean IoU (%): 48.10% -> 68.40% | Improved |
| propagate change by geometric consistency instead of photoconsistency | Taneja et al. · 3RScan change detection | mean recall (%): 44.50% -> 76.05% | Improved |
| use ground-truth transformations instead of computed ones (ablation) | Ours · 3RScan change detection | mean IoU (%): 68.40% -> 72.40% | Improved |
| use ground-truth transformations instead of computed ones (ablation) | Ours · 3RScan change detection | mean recall (%): 76.05% -> 93.89% | Improved |
| add Mask-RCNN semantic labels in place of geometric consistency (ablation) | Ours · 3RScan change detection | mean IoU (%): 68.40% -> 52.96% | Hurt |
| add Mask-RCNN semantic labels in place of geometric consistency (ablation) | Ours · 3RScan change detection | mean recall (%): 76.05% -> 89.22% | Improved |
| use only RANSAC inliers without initial detection or optimization (ablation) | Ours · 3RScan change detection | mean IoU (%): 68.40% -> 10.82% | Hurt |
| increase number of top transformations k from 5 to 10 (ablation) | Ours · 3RScan change detection | mean IoU (%): 68.40% -> 64.89% | No clear effect |
| increase number of top transformations k from 5 to 15 (ablation) | Ours · 3RScan change detection | mean recall (%): 76.05% -> 79.09% | Mixed |
| add geometric graph cut optimization (ablation) | Palazzolo et al. · 3RScan change detection | mean accuracy (%): 67.76% -> 54.60% | Hurt |
Closest papers by meaning
- Multi-View Pose-Agnostic Change Localization with Zero Labels (CVPR 2025 · similarity 0.87)
- Living Scenes: Multi-object Relocalization and Reconstruction in Changing 3D Environments (CVPR 2024 · similarity 0.85)
- Tracking and Understanding Object Transformations (NeurIPS 2025 · similarity 0.84)
- Stereo4D: Learning How Things Move in 3D from Internet Stereo Videos (CVPR 2025 · similarity 0.84)
- M^3-VOS: Multi-Phase, Multi-Transition, and Multi-Scenery Video Object Segmentation (CVPR 2025 · similarity 0.84)
- Continuous 3D Perception Model with Persistent State (CVPR 2025 · 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.
