Research map / Detection and segmentation / Open-set and few-shot detection
O-MaMa: Learning Object Mask Matching between Egocentric and Exocentric Views
Lorenzo Mur-Labadia, Maria Santos-Villafranca, Jesus Bermudez-Cameo, Alejandro Perez-Yus, Ruben Martinez-Cantin, Jose J. Guerrero
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 Mask Matching contrastive loss with an MLP (ablation) | k-NN baseline · Ego-Exo4D val 10% Ego2Exo | IoU: 35.2 -> 42.2 | Improved |
| add Mask Matching contrastive loss with an MLP (ablation) | k-NN baseline · Ego-Exo4D val 10% Exo2Ego | IoU: 34.9 -> 44.7 | Improved |
| add context descriptor without adjacent negatives (ablation) | O-MaMa (loss only) · Ego-Exo4D val 10% Ego2Exo | IoU: 42.2 -> 42.7 | No clear effect |
| add adjacent hard negative mining (ablation) | O-MaMa + context · Ego-Exo4D val 10% Ego2Exo | IoU: 42.7 -> 46.9 | Improved |
| add Ego-Exo cross attention (weighted-sum union) (ablation) | O-MaMa + adjacent negatives · Ego-Exo4D val 10% Ego2Exo | IoU: 46.9 -> 47.3 | Improved |
| use concat instead of weighted sum for global union (ablation) | O-MaMa · Ego-Exo4D val 10% Ego2Exo | IoU: 47.3 -> 48.3 | Improved |
| add epipolar geometry constraint to k-NN (ablation) | k-NN Avg-Pool(mask) DINOv2 · Ego-Exo4D val 10% | IoU: 35.2 vs. 35.4 | No clear effect |
| average DINOv2 mask features instead of CLIP bbox pooling (ablation) | k-NN baseline · Ego-Exo4D val 10% Ego2Exo | IoU: 24.5 -> 35.2 | Improved |
| pool DINOv2 over mask instead of bounding box (ablation) | k-NN baseline · Ego-Exo4D val 10% Ego2Exo | IoU: 21.8 -> 35.2 | Improved |
| pool DINOv2 over mask instead of mask centroid (ablation) | k-NN baseline · Ego-Exo4D val 10% Ego2Exo | IoU: 25.6 -> 35.2 | Improved |
| full O-MaMa modules vs k-NN baseline | O-MaMa · Ego-Exo4D val 10% Ego2Exo | Loc.E: +67.5 | Improved |
| reformulate as mask matching vs zero-shot PSALM | O-MaMa k-NN baseline · Ego-Exo4D v1 val Ego2Exo | IoU: 7.9 -> 40.5 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- ObjectRelator — ObjectRelator: Enabling Cross-View Object Relation Understanding Across Ego-Centric and Exo-Centric Perspectives (ICCV 2025) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Segment Anything, Even Occluded (CVPR 2025 · similarity 0.87)
- XMask3D: Cross-modal Mask Reasoning for Open Vocabulary 3D Semantic Segmentation (NeurIPS 2024 · similarity 0.87)
- v-CLR: View-Consistent Learning for Open-World Instance Segmentation (CVPR 2025 · similarity 0.86)
- EAGLE: Efficient Adaptive Geometry-based Learning in Cross-view Understanding (NeurIPS 2024 · similarity 0.86)
- BLADE: Box-Level Supervised Amodal Segmentation through Directed Expansion (AAAI 2024 · similarity 0.86)
- Mask-Adapter: The Devil is in the Masks for Open-Vocabulary Segmentation (CVPR 2025 · similarity 0.85)
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.
