Research map / Detection and segmentation / Open-set and few-shot detection
Functionality Discovery and Prediction of Physical Objects
Lei Ji, Botian Shi, Xianglin Guo, Xilin Chen
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 consolidation of ConceptNet, pattern and model-based tuples | Head-Action extraction · Head-Action dataset | F1: 0.580 -> 0.610 | Improved |
| use consolidation instead of ConceptNet only | Head-Action extraction · Head-Action dataset | recall: 0.152 -> 0.624 | Improved |
| use consolidation instead of ConceptNet only | Head-Action extraction · Head-Action dataset | precision: 0.862 -> 0.597 | Hurt |
| use model-based instead of pattern-based extraction | Head-Action extraction · Head-Action dataset | precision: 0.632 -> 0.694 | Improved |
| use model-based instead of pattern-based extraction | Head-Action extraction · Head-Action dataset | recall: 0.536 -> 0.255 | Hurt |
| directly classify functionality from image instead of object classification + PRA | object-functionality prediction · Head(image)-Action-Tail test | F1: 0.486 -> 0.756 | Improved |
| use combined PRA+Dismult inference after object classification instead of functionality classification | object-functionality prediction · Head(image)-Action-Tail test | F1: 0.756 -> 0.7789 | Improved |
| combine functionality classification with combined inference scores | object-functionality prediction · Head(image)-Action-Tail test | F1: 0.7789 -> 0.810 | Improved |
| use Dismult instead of PRA after object classification | object-functionality prediction · Head(image)-Action-Tail test | recall: 0.388 -> 0.633 | Improved |
| add object classification to functionality classification (joint model) (ablation) | visual module · functionality classification | F1: 0.734 -> 0.731 | No clear effect |
| train Dismult on consolidated graph instead of ConceptNet (ablation) | Dismult module · object-functionality prediction | F1: 0.539 -> 0.651 | Improved |
| use consolidated graph instead of ConceptNet for PRA (ablation) | PRA module · random sample | F1: 0.134 -> 0.552 | Improved |
Closest papers by meaning
- Joint Discovery of Object States and Manipulation Actions (ICCV 2017 · similarity 0.84)
- Functionality Understanding and Segmentation in 3D Scenes (CVPR 2025 · similarity 0.83)
- Unsupervised Learning for Physical Interaction through Video Prediction (NeurIPS 2016 · similarity 0.82)
- Understanding 3D Object Interaction from a Single Image (ICCV 2023 · similarity 0.82)
- InfoCon: Concept Discovery with Generative and Discriminative Informativeness (ICLR 2024 · similarity 0.82)
- RobotSmith: Generative Robotic Tool Design for Acquisition of Complex Manipulation Skills (NeurIPS 2025 · similarity 0.80)
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.
