[TPAMI-2023] Dataset-Driven Unsupervised Object Discovery for Region-Based Instance Image Retrieval

论文阅读 [TPAMI-2023] Dataset-Driven Unsupervised Object Discovery for Region-Based Instance Image Retrieval

论文搜索(studyai.com)

搜索论文: Dataset-Driven Unsupervised Object Discovery for Region-Based Instance Image Retrieval

搜索论文: http://www.studyai.com/search/whole-site/?q=Dataset-Driven+Unsupervised+Object+Discovery+for+Region-Based+Instance+Image+Retrieval&fr=csdn

关键字(Keywords)

Detectors; Image retrieval; Feature extraction; Object detection; Task analysis; Reliability; Training; Instance image retrieval; region-based retrieval; weakly-supervised object detection; unsupervised learning

机器学习; 机器视觉

无监督学习; 弱监督学习; 目标检测; 图像检索

摘要(Abstract)

Instance image retrieval could greatly benefit from discovering objects in the image dataset.

实例图像检索可以从发现图像数据集中的对象中获益匪浅。

This not only helps produce more reliable feature representation but also better informs users by delineating query-matched object regions.

这不仅有助于产生更可靠的特征表示，而且通过描绘查询匹配的对象区域来更好地通知用户。

However, object classes are usually not predefined in a retrieval dataset and class label information is generally unavailable in image retrieval.

然而，对象类通常不会在检索数据集中预定义，并且类标签信息通常在图像检索中不可用。

This situation makes object discovery a challenging task.

这种情况使得对象发现成为一项具有挑战性的任务。

To address this, we propose a novel dataset-driven unsupervised object discovery framework.

为了解决这个问题，我们提出了一种新的数据集驱动的无监督对象发现框架。

By utilizing deep feature representation and weakly-supervised object detection, we explore supervisory information from within an image dataset, construct class-wise object detectors, and assign multiple detectors to each image for detection.

通过利用深度特征表示和弱监督对象检测，我们从图像数据集中探索监督信息，构建类对象检测器，并为每个图像分配多个检测器进行检测。

To efficiently construct object detectors for large image datasets, we propose a novel “base-detector repository” and derive a fast way to generate the base detectors.

为了有效地构建大型图像数据集的目标检测器，我们提出了一种新的“基础检测器库”，并导出了一种快速生成基础检测器的方法。

In addition, the whole framework is designed to work in a self-boosting manner to iteratively refine object discovery.

此外，整个框架被设计为以自增强的方式工作，以迭代优化对象发现。

Compared with existing unsupervised object detection methods, our framework produces more accurate object discovery results.

与现有的无监督对象检测方法相比，我们的框架产生了更准确的对象发现结果。

Different from supervised detection, we need neither manual annotation nor auxiliary datasets to train object detectors.

与监督检测不同，我们既不需要手动注释，也不需要辅助数据集来训练对象检测器。

Experimental study demonstrates the effectiveness of the proposed framework and the improved performance for region-based instance image retrieval…

实验研究证明了所提出的框架的有效性，并提高了基于区域的实例图像检索的性能。

作者(Authors)

[‘Zhongyan Zhang’, ‘Lei Wang’, ‘Yang Wang’, ‘Luping Zhou’, ‘Jianjia Zhang’, ‘Fang Chen’]