Research map / Detection and segmentation / Open-set and few-shot detection
Object Pursuit: Building a Space of Objects via Discriminative Weight Generation
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Use the learned base-object manifold instead of searching the full representation space | Learning a new object representation with Object Pursuit · iThor, 1-shot | segmentation accuracy: 0.225 -> 0.288; 57.6% | Helped |
| Use the learned base-object manifold instead of searching the full representation space | Learning a new object representation with Object Pursuit · iThor, 5-shot | segmentation accuracy: 0.468 -> 0.634; 35.5% | Helped |
| Use the learned base-object manifold instead of searching the full representation space | Learning a new object representation with Object Pursuit · iThor, 10-shot | segmentation accuracy: 0.524 -> 0.677; 29.2% | Helped |
| Increase the quality threshold τ (ablation) | Object Pursuit · iThor pursuit | average segmentation accuracy on learned objects: 0.75 -> 0.77 -> 0.83 -> 0.86 | Helped |
| Increase τ to 0.8 rather than use τ=0.7 (ablation) | Learning new-object representations on the base manifold · iThor, 2000-shot | segmentation accuracy: 0.740 -> 0.731 | Hurt |
| Remove the forgetting-prevention constraint (ablation) | Object Pursuit · iThor pursuit | average segmentation accuracy on learned objects: 0.71 -> 0.02 | Hurt |
| Remove the forgetting-prevention constraint (ablation) | Object Pursuit · iThor pursuit | forgetting rate: 0.02 -> 0.97 | Hurt |
| Increase the sparsity constraint on μ from |α|=0.0 to |α|=0.2 (ablation) | Object Pursuit manifold-expression test · iThor pursuit | rate of objects expressed by bases: 0.11 -> 0.21 | Helped |
| Increase the sparsity constraint on μ to |α|=0.5 (ablation) | Object Pursuit manifold-expression test · iThor pursuit | rate of objects expressed by bases: 0.21 -> 0.08 | Hurt |
| Increase the sparsity constraint on z (ablation) | Object Pursuit · iThor pursuit | rate of objects expressed by bases: 0.12 -> 0.21 -> 0.21 -> 0.24 | Helped |
| Increase the sparsity constraint on z to |α|=0.5 (ablation) | Object Pursuit · iThor pursuit | portion of learnable objects added to the object list: 0.58 -> 0.55 | Hurt |
| Raise τ from 0.5 to 0.8 (ablation) | Object Pursuit re-identification · Unseen test objects | recall: 0.28 -> 0.72 | Helped |
Closest papers by meaning
- Temporally Consistent Object-Centric Learning by Contrasting Slots (CVPR 2025 · similarity 0.89)
- Object-Centric Pretraining via Target Encoder Bootstrapping (ICLR 2025 · similarity 0.89)
- Learning to Compose: Improving Object Centric Learning by Injecting Compositionality (ICLR 2024 · similarity 0.88)
- Multi-Part Object Representations via Graph Structures and Co-Part Discovery (NeurIPS 2026 · similarity 0.88)
- Object Concepts Emerge from Motion (NeurIPS 2025 · similarity 0.88)
- Object centric architectures enable efficient causal representation learning (ICLR 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.