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 2404.01524cs.CVcs.AI
收藏

关于图像检索中训练-测试类别重叠及其检测的研究

On Train-Test Class Overlap and Detection for Image Retrieval

Dealicious Inc.(Dealicious公司);Sejong University(世宗大学);Institute of Advanced Research on Artificial Intelligence (IARAI)(人工智能高级研究院(IARAI))

Chull Hwan Song, Jooyoung Yoon, Taebaek Hwang, Shunghyun Choi, Yeong Hyeon Gu, Yannis Avrithis

AI总结:

该研究揭示图像检索训练集与测试集类别重叠会显著且不一致地抬高性能,并提出端到端单阶段CiDeR框架,联合检测目标与学习全局表征,在清理后的RGLDv2-clean等基准上取得最优结果。

摘要:

在图像检索中,训练集与评估集之间不存在类别重叠究竟有多重要?我们重新审视了最常用的训练集 Google Landmarks v2 clean,通过识别并移除其与最常用评估集 Revisited Oxford and Paris [34] 之间的类别重叠。通过在复现的最先进方法基准上比较原始数据集与新构建的 RGLDv2-clean,我们的发现令人震惊:不仅性能出现显著下降,而且这种下降在不同方法之间并不一致,从而改变了方法排名。在构建索引时,要专注于感兴趣的对象并忽略背景杂乱需要什么条件?我们是否需要分别训练目标检测器和表示模型?是否需要位置监督?我们提出了 Single-stage Detect-to-Retrieve(CiDeR),这是一个端到端的单阶段流程,用于检测感兴趣的对象并提取全局图像表示。我们在现有训练集和新构建的 RGLDv2-clean 上均超越了此前的最先进方法。我们的数据集可在 https://github.com/dealicious-inc/RGLDv2-clean 获取。

英文摘要:

How important is it for training and evaluation sets to not have class overlap in image retrieval? We revisit Google Landmarks v2 clean, the most popular training set, by identifying and removing class overlap with Revisited Oxford and Paris [34], the most popular evaluation set. By comparing the original and the new RGLDv2-clean on a benchmark of reproduced state-of-the-art methods, our findings are striking. Not only is there a dramatic drop in performance, but it is inconsistent across methods, changing the ranking.What does it take to focus on objects or interest and ignore background clutter when indexing? Do we need to train an object detector and the representation separately? Do we need location supervision? We introduce Single-stage Detect-to-Retrieve (CiDeR), an end-to-end, single-stage pipeline to detect objects of interest and extract a global image representation. We outperform previous state-of-the-art on both existing training sets and the new RGLDv2-clean. Our dataset is available at https://github.com/dealicious-inc/RGLDv2-clean.

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。

↑