Research map / Detection and segmentation / Open-set and few-shot detection
PONI: Potential Functions for ObjectGoal Navigation With Interaction-Free Learning
Check 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 |
|---|---|---|---|
| replace SemExp's interactive long-term-goal learning with PONI's interaction-free potential prediction | ObjectNav · Gibson (val) | SPL: 39.6 -> 41.0 | Helped |
| use PONI rather than THDA | ObjectNav · MP3D (val) | SPL: 11.0 -> 12.1 | Helped |
| predict area and object potential functions instead of directly predicting object direction or location | PONI's semantic-map policy · Gibson (val) | Success: 69.9 -> 73.6 | Helped |
| add the area potential to the object potential (ablation) | PONI · Gibson (val) | Success: 58.8 -> 73.6 | Helped |
| add the area potential to the object potential (ablation) | PONI · MP3D (val) | Success: 30.5 -> 31.8 | Helped |
| add the object potential to area-potential exploration (ablation) | PONI · Gibson (val) | Success: 65.1 -> 73.6 | Helped |
| add the object potential to area-potential exploration (ablation) | PONI · MP3D (val) | Success: 30.8 -> 31.8 | Helped |
| define both potentials only at map frontiers (ablation) | PONI · Gibson (val) | Success: 72.7 -> 73.6 | Helped |
| replace predicted image segmentation with ground-truth segmentation (ablation) | PONI · Gibson (val) | Success: 73.6 -> 86.5 | Helped |
| replace predicted image segmentation with ground-truth segmentation (ablation) | PONI · MP3D (val) | Success: 31.8 -> 58.2 | Helped |
| use view-cone rather than square masks to generate potential-network training samples (ablation) | PONI · MP3D (val) | SPL: 12.1 -> 12.1 | No clear effect |
| add distance-based action cost when sampling long-term goals (ablation) | PONI · MP3D (val) | Success: 31.8 -> 30.3 | Hurt |
Closest papers by meaning
- Object Goal Navigation using Goal-Oriented Semantic Exploration (NeurIPS 2020 · similarity 0.87)
- ESC: Exploration with Soft Commonsense Constraints for Zero-shot Object Navigation (ICML 2023 · similarity 0.87)
- Imagine Before Go: Self-Supervised Generative Map for Object Goal Navigation (CVPR 2024 · similarity 0.86)
- An Interactive Navigation Method with Effect-oriented Affordance (CVPR 2024 · similarity 0.86)
- Trajectory Diffusion for ObjectGoal Navigation (NeurIPS 2024 · similarity 0.85)
- BOP-ASK: Object-Interaction Reasoning for Vision-Language Models (CVPR 2026 · 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.