CGFormer

Datasets for open-vocabulary settings, including the RefCOCO series, were utilized for **CGFormer (CVPR 2023): Contrastive Grouping with Transformer for Referring Image Segmentation.** This paper first introduces learnable query tokens to represent objects and then alternately queries linguistic features and groups visual features into the query tokens for object-aware cross-modal reasoning. CGFormer achieves cross-level interaction by jointly updating the query tokens and decoding masks in every two consecutive layers. In addition, we introduce new splits on datasets for evaluating generalization for referring image segmentation models. ### Download from modelscope.msdatasets import MsDataset ds = MsDataset.load('Toneyaya/CGFormer') or modelscope download --dataset Toneyaya/CGFormer本研究将适用于开放词汇场景（open-vocabulary settings）的数据集（包含RefCOCO系列数据集）应用于**CGFormer（CVPR 2023：面向指代表象分割的Transformer对比分组方法）**。 该论文首先提出可学习查询Token（learnable query tokens）以表征目标对象，随后交替查询语言特征并将视觉特征聚合至查询Token中，实现面向目标感知的跨模态推理。CGFormer通过在每连续两层中同步更新查询Token与解码掩码，实现跨层级交互。此外，本研究还构建了数据集的全新划分方案，用于评估指代表象分割模型的泛化性能。 ### 下载 python from modelscope.msdatasets import MsDataset ds = MsDataset.load('Toneyaya/CGFormer') 或者 bash modelscope download --dataset Toneyaya/CGFormer

搜集汇总

背景与挑战

背景概述

该数据集用于开放词汇设置下的引用图像分割研究，包含RefCOCO系列数据，支持CVPR 2023论文CGFormer的方法。它引入了新的数据集分割，旨在评估模型的泛化能力。