PropVG-Data
收藏资源简介：
PropVG视觉定位数据集包含了用于PropVG模型的标注数据，该模型是一种端到端的提议驱动视觉定位框架，适用于无需额外检测器的情况下，将前景对象提议生成与参照对象理解无缝集成。数据集涵盖了多个视觉定位基准，如RefCOCO、gRefCOCO等，包含了相应的参照标注和前景标注文件。
The PropVG Visual Grounding Dataset contains annotated data for the PropVG model, which is an end-to-end proposal-driven visual grounding framework that seamlessly integrates foreground object proposal generation and referring object understanding without requiring additional detectors. The dataset covers multiple visual grounding benchmarks such as RefCOCO, gRefCOCO, etc., and includes corresponding referring annotations and foreground annotation files.
PropVG Visual Grounding Datasets 概述
数据集基本信息
- 许可证: Apache-2.0
- 任务类别: 零样本目标检测
- 语言: 英语
- 标签: 视觉定位、目标检测、指代表达式理解
数据集来源
该数据集为PropVG模型使用的标注数据，相关论文为《PropVG: End-to-End Proposal-Driven Visual Grounding with Multi-Granularity Discrimination》。
数据集结构
数据集包含多个视觉定位基准的标注文件，需单独准备相应图像数据集（如MSCOCO）。
目录结构
| -- data |--seqtr_type | -- annotations | -- refcoco-unc | -- instances_withid.json | -- refcocoplus-unc | -- instances_withid.json | -- refcocog-umd | -- instances_withid.json | -- mixed-seg | -- instances_nogoogle_withid.json | -- coco_all.json | -- grefs | -- instance.json | -- coco_annotations.json | -- ref-zom | -- instances_withid.json | -- allobj.json | -- rrefcoco | -- instances_withid.json | -- allobj.json | -- images | -- mscoco | -- train2014
标注文件详情
| 训练集 | 指代表注文件 | 前景标注文件 |
|---|---|---|
| RefCOCO | instances_withid.json | coco_all.json |
| RefCOCO+ | instances_withid.json | coco_all.json |
| RefCOCOg | instances_withid.json | coco_all.json |
| RefCOCO-mix | instances_nogoogle_withid.json | coco_all.json |
| gRefCOCO | instances_withid.json | coco_annotations.json |
| Ref-ZOM | instances_withid.json | allobj.json |
| RRefCOCO | instances_withid.json | allobj.json |
使用示例
GRES任务演示
bash python tools/demo.py --img "asserts/imgs/Figure_1.jpg" --expression "three skateboard guys" --config "configs/gres/PropVG-grefcoco.py" --checkpoint /PATH/TO/PropVG-grefcoco.pth --img_size 320
RIS任务演示
bash python tools/demo.py --img "asserts/imgs/Figure_2.jpg" --expression "full half fruit" --config "configs/refcoco/PropVG-refcoco-mix.py" --checkpoint /PATH/TO/PropVG-refcoco-mix.pth --img_size 384
引用信息
bibtex @misc{propvg, title={PropVG: End-to-End Proposal-Driven Visual Grounding with Multi-Granularity Discrimination}, author={Ming Dai and Wenxuan Cheng and Jiedong Zhuang and Jiang-jiang Liu and Hongshen Zhao and Zhenhua Feng and Wankou Yang}, year={2025}, eprint={2509.04833}, archivePrefix={arXiv}, primaryClass={cs.CV}, url={https://arxiv.org/abs/2509.04833}, }
