Research map / Detection and segmentation / Open-set and few-shot detection
MarvelOVD: Marrying Object Recognition and Vision-Language Models for Robust Open-Vocabulary Object Detection
Kuo Wang, Lechao Cheng, Weikai Chen, Pingping Zhang, Liang Lin, Fan Zhou, Guanbin Li
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 online object mining (detector novelty score combined with CLIP score) (ablation) | VL-PLM pipeline · COCO OVD | AP50 novel: 34.4 -> 37.8 | Improved |
| add adaptive proposal reweighting (ablation) | MarvelOVD · COCO OVD | AP50 novel: 37.8 -> 38.9 | Improved |
| add stratified label assignment (ablation) | VL-PLM with weak-strong augmentation · COCO OVD | AP50 base: 53.9 -> 56.4 | Improved |
| add weak-strong augmentation (ablation) | VL-PLM · COCO OVD | AP50 novel: 32.5 -> 34.2 | Improved |
| increase novel loss weight gamma from 1 to 2 (ablation) | VL-PLM · COCO OVD | AP50 novel: 32.7 -> 32.5 | No clear effect |
| increase novel loss weight gamma from 1 to 2 (ablation) | MarvelOVD · COCO OVD | AP50 novel: 37.8 -> 38.9 | Improved |
| use background-score reliability (1-b_i) instead of pseudo-label confidence (ablation) | adaptive reweighting · COCO OVD | AP50 novel: 37.8 -> 39.8 | Improved |
| lengthen burn-in from 0.5k to 5k steps (ablation) | MarvelOVD · COCO OVD | AP50 novel: 38.9 -> 38.5 | No clear effect |
| set lambda to 0.5 mixing CLIP and detector scores rather than extremes (ablation) | online object mining · COCO OVD | AP50 novel | Improved |
| raise pseudo-label threshold from 0.8 to 0.9 (ablation) | MarvelOVD · COCO OVD | AP50 novel: 37.0 -> 38.9 | Improved |
Closest papers by meaning
- Taming Self-Training for Open-Vocabulary Object Detection (CVPR 2024 · similarity 0.93)
- Language-conditioned Detection Transformer (CVPR 2024 · similarity 0.92)
- How to Evaluate the Generalization of Detection? A Benchmark for Comprehensive Open-Vocabulary Detection (AAAI 2024 · similarity 0.91)
- LLMs Meet VLMs: Boost Open Vocabulary Object Detection with Fine-grained Descriptors (ICLR 2024 · similarity 0.91)
- DitHub: A Modular Framework for Incremental Open-Vocabulary Object Detection (NeurIPS 2025 · similarity 0.90)
- Roboflow100-VL: A Multi-Domain Object Detection Benchmark for Vision-Language Models (NeurIPS 2025 · similarity 0.90)
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.
