Research map / Detection and segmentation / Open-set and few-shot detection
Vision-Language Fusion for Object Recognition
Sz-Rung Shiang, Stephanie Rosenthal, Anatole Gershman, Jaime Carbonell, Jean Oh
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 10 spatial relations per image | MultiRank · ground-truth boxes | accuracy: 0.6299 -> 0.7240 | Improved |
| add 10 spatial relations per image | MultiRank · ground-truth boxes | mAP: 0.7240 -> 0.8002 | Improved |
| add only 1 spatial relation (ablation) | MultiRank · ground-truth boxes | accuracy: 0.6299 -> 0.6331 | No clear effect |
| increase number of relations from 1 to 10 (ablation) | MultiRank · ground-truth boxes | accuracy: 0.6331 -> 0.7240 | Improved |
| use naive fusion instead of MultiRank with 10 relations (ablation) | naive fusion · ground-truth boxes | accuracy: 0.7240 -> 0.6527 | Hurt |
| use confusion matrix plus co-occurrence without human descriptions | MultiRank · ground-truth boxes | accuracy: 0.6299 -> 0.6558 | Improved |
| use co-occurrence only (ablation) | MultiRank · ground-truth boxes | accuracy: 0.6299 -> 0.6331 | No clear effect |
| use detected instead of ground-truth boxes for vision-only | vision-only · detected boxes | accuracy: 62.99% to 42.28% | Hurt |
| add 10 relations on detected boxes | MultiRank · detected boxes | accuracy: 0.4229 -> 0.5885 | Improved |
| choose relations on error-prone boxes (10 relations, detected boxes) | MultiRank · detected boxes | accuracy: 0.5885 -> 0.6857 | Improved |
| choose relations on error-prone boxes (1 relation, ground-truth boxes) | MultiRank · ground-truth boxes | accuracy: 0.6331 -> 0.6526 | Improved |
| choose relations on error-prone boxes (10 relations, ground-truth boxes) | MultiRank · ground-truth boxes | accuracy: 0.7240 -> 0.7241 | No clear effect |
Closest papers by meaning
- ORIC: Benchmarking Object Recognition under Contextual Incongruity in Large Vision-Language Models (CVPR 2026 · similarity 0.86)
- Synthesize Diagnose and Optimize: Towards Fine-Grained Vision-Language Understanding (CVPR 2024 · similarity 0.85)
- Roboflow100-VL: A Multi-Domain Object Detection Benchmark for Vision-Language Models (NeurIPS 2025 · similarity 0.85)
- Object Recognition as Next Token Prediction (CVPR 2024 · similarity 0.84)
- Image Fusion via Vision-Language Model (ICML 2024 · similarity 0.84)
- Reading Images Like Texts: Sequential Image Understanding in Vision-Language Models (ICLR 2026 · 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.
