Research map / Video / Video segmentation and tracking
Instance-Level Moving Object Segmentation from a Single Image with Events
Zhexiong Wan, Bin Fan, Le Hui, Yuchao Dai, Gim Hee Lee
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 cross-modal masked attention (CMA) (ablation) | basic InsMOS model · EKubric | mAP: 32.4 -> 36.4 | Improved |
| add flow-guided feature enhancement (FFE) (ablation) | CMA model · EKubric | mAP: 36.4 -> 38.5 | Improved |
| add contrastive feature learning (CFL) on top of CMA+FFE (ablation) | CMA+FFE · EKubric | mAP: 42.8 -> 44.6 | Improved |
| add events to a single image instead of two images (ablation) | complete model · EKubric | mAP: 35.4 -> 44.6 | Improved |
| use events instead of precomputed RAFT flow as motion source (ablation) | complete model · EKubric | mAP: 34.8 -> 44.6 | Improved |
| add image to events-only input (ablation) | complete model · EKubric | mAP: 32.9 -> 44.6 | Improved |
| augment both texture and motion in CMA instead of one direction (ablation) | complete model · EKubric | mAP: 40.9 -> 44.6 | Improved |
| add cross-similarity term in contrastive loss (ablation) | complete model · EKubric | mAP: 43.7 -> 44.6 | Improved |
| increase training frames K from 2 to 3 or 4 (ablation) | complete model · EKubric | mAP: 44.6 -> 45.0 | No clear effect |
| reduce batch size from 16 to 4 (ablation) | complete model · EKubric | mAP: 44.6 -> 44.1 | Hurt |
| raise moving threshold from 0.3 to 0.7 (ablation) | complete model · EKubric | mIoU_ins: 79.66 -> 71.70 | Hurt |
| lower moving threshold from 0.3 to 0.1 (ablation) | complete model · EKubric | mAP: 44.6 -> 38.7 | Hurt |
Closest papers by meaning
- DIMOS: Disentangling Instance-level Moving Object Segmentation (CVPR 2026 · similarity 0.92)
- Event-Based Motion Segmentation by Motion Compensation (ICCV 2019 · similarity 0.92)
- Segment Any Motion in Videos (CVPR 2025 · similarity 0.91)
- Know Where You Are From: Event-Based Segmentation via Spatio-Temporal Propagation (AAAI 2025 · similarity 0.88)
- TrackIME: Enhanced Video Point Tracking via Instance Motion Estimation (NeurIPS 2024 · similarity 0.88)
- Learning Segmentation from Point Trajectories (NeurIPS 2024 · similarity 0.88)
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.
