arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2407.02389cs.CVcs.AIcs.CLcs.LGcs.MM
收藏

SafaRi:用于弱监督指代表达分割的自适应序列 Transformer

SafaRi:Adaptive Sequence Transformer for Weakly Supervised Referring Expression Segmentation

University of Toronto(多伦多大学);Adobe Research(奥多比研究院)

Sayan Nag, Koustava Goswami, Srikrishna Karanam

AI总结:

提出弱监督 RES 架构 SafaRi,通过注意力一致的跨模态融合与空间感知零样本伪标签过滤,仅用30%标注即超过全监督 SeqTR,并具备零样本泛化能力。

摘要:

指代表达分割(RES)旨在为图像中由文本(即指代表达)所指代的目标对象提供分割掩码。现有方法需要大规模掩码标注,而且这类方法不能很好地泛化到未见/零样本场景。为解决上述问题,我们提出了一种用于 RES 的弱监督自举架构,并包含若干新的算法创新。据我们所知,我们的方法是首个在训练中仅考虑一部分掩码标注和框标注的方法(如图 1 和表 1 所示)。为了在这种低标注设置下实现有原则的模型训练、改进图文区域级对齐,并进一步增强图像中目标对象的空间定位,我们提出了带注意力一致性的跨模态融合模块。对于未标注样本的自动伪标注,我们引入了一种新的掩码有效性过滤流程,该流程基于具有空间感知能力的零样本候选框评分方法。大量实验表明,仅使用 30% 的标注,我们的模型 SafaRi 在 RefCOCO+@testA 和 RefCOCO+testB 数据集上分别取得 59.31 和 48.26 mIoU,而全监督 SOTA 方法 SeqTR 分别取得 58.93 和 48.19 mIoU。在全监督设置下,SafaRi 还分别在 RefCOCO+testA 和 RefCOCO+testB 上比 SeqTR 高出 11.7% 和 19.6%,并在未见/零样本任务中展现出强大的泛化能力。

英文摘要:

Referring Expression Segmentation (RES) aims to provide a segmentation mask of the target object in an image referred to by the text (i.e., referring expression). Existing methods require large-scale mask annotations. Moreover, such approaches do not generalize well to unseen/zero-shot scenarios. To address the aforementioned issues, we propose a weakly-supervised bootstrapping architecture for RES with several new algorithmic innovations. To the best of our knowledge, ours is the first approach that considers only a fraction of both mask and box annotations (shown in Figure 1 and Table 1) for training. To enable principled training of models in such low-annotation settings, improve image-text region-level alignment, and further enhance spatial localization of the target object in the image, we propose Cross-modal Fusion with Attention Consistency module. For automatic pseudo-labeling of unlabeled samples, we introduce a novel Mask Validity Filtering routine based on a spatially aware zero-shot proposal scoring approach. Extensive experiments show that with just 30% annotations, our model SafaRi achieves 59.31 and 48.26 mIoUs as compared to 58.93 and 48.19 mIoUs obtained by the fully-supervised SOTA method SeqTR respectively on RefCOCO+@testA and RefCOCO+testB datasets. SafaRi also outperforms SeqTR by 11.7% (on RefCOCO+testA) and 19.6% (on RefCOCO+testB) in a fully-supervised setting and demonstrates strong generalization capabilities in unseen/zero-shot tasks.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑