Research map / Detection and segmentation / Open-set and few-shot detection
Task-driven Processing with Coarse-to-Fine Glimpse-based Active Perception
Oleh Kolner, Thomas Ortner, Stanislaw Wozniak, Angeliki Pantazi
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 CF-GAP front-end to NIDS-Net | NIDS-Net · HR-InsDet | AP (avg): 63.8 -> 73.3 | Improved |
| add CF-GAP front-end to OTS-FM MobileSAM | OTS-FM MobileSAM · HR-InsDet hard subset | AP: 22.0 -> 41.7 | Improved |
| add CF-GAP front-end to OTS-FM SAM | OTS-FM SAM · HR-InsDet | AP (avg): 41.6 -> 56.5 | Improved |
| add CF-GAP front-end to OTS-FM GroundingDINO | OTS-FM GroundingDINO · HR-InsDet small subset | AP: 28.8 -> 39.6 | Improved |
| add CF-GAP front-end to NIDS-Net | NIDS-Net · Robotools | AP: 64.9 -> 70.3 | Improved |
| add CF-GAP front-end to OTS-FM MobileSAM | OTS-FM MobileSAM · Robotools | AP: 35.5 -> 49.4 | Improved |
| add fine glimpsing (ablation) | CF-GAP + NIDS-Net · HR-InsDet hard subset | AP: 56.2 -> 63.2 | Improved |
| add fine glimpsing (ablation) | CF-GAP + NIDS-Net · HR-InsDet large subset | AP: 85.5 -> 86.1 | No clear effect |
| provide glimpse locations to downstream architecture (ablation) | CF-GAP + NIDS-Net · HR-InsDet hard subset | AP: 49.5 -> 63.2 | Improved |
| use attention-based encoders instead of CNN (ablation) | CF-GAP + NIDS-Net · HR-InsDet hard subset | AP: 58.6 -> 63.2 | Improved |
| use CF-GAP instead of exhaustive patching | NIDS-Net · HR-InsDet hard subset | AP: 49.1 -> 63.2 | Improved |
| patch scene into RoIs for OTS-FM MobileSAM | OTS-FM MobileSAM · HR-InsDet hard subset | AP: 22.0 -> 23.2 | No clear effect |
Closest papers by meaning
- CanViT: Toward Active-Vision Foundation Models (NeurIPS 2026 · similarity 0.88)
- Mind the GAP: Glimpse-based Active Perception improves generalization and sample efficiency of visual reasoning (ICLR 2025 · similarity 0.87)
- LookWhere? Efficient Visual Recognition by Learning Where to Look and What to See from Self-Supervision (NeurIPS 2025 · similarity 0.84)
- ActiveScope: Actively Seeking and Correcting Perception for MLLMs (ICML 2026 · similarity 0.84)
- I-Perceive: A Foundation Model for Vision-Language Active Perception (NeurIPS 2026 · similarity 0.84)
- Seeing More with Less: Human-like Representations in Vision Models (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.
