一种新的无监督前景目标检测方法 A New Unsupervised Foreground Object Detection Method

14.一种新的无监督前景目标检测方法
A New Unsupervised Foreground Object Detection Method
摘要：针对基于无监督特征提取的目标检测方法效率不高的问题,提出一种在无标记数据集中准确检测前景目标的方法.其基本出发点是:正确的特征聚类结果可以指导目标特征提取,同时准确提取的目标特征可以提高特征聚类的精度.该方法首先对无标记样本图像进行局部特征提取,然后根据最小化特征距离进行无监督特征聚类.将同一个聚类内的图像两两匹配,将特征匹配的重现程度作为特征权重,最后根据更新后的特征权重指导下一次迭代的特征聚类.多次迭代后同时得到聚类结果和前景目标.实验结果表明,该方法有效地提高Caltech256数据集和Google车辆图像的检测精度.此外,针对目前绝大部分无监督目标检测方法不具备增量学习能力这一缺点,提出了增量学习方法实现,实验结果表明,增量学习方法有效地提高了计算速度.

关键词: 前景目标检测, 无监督学习, 特征聚类, 增量学习, 特征提取

Abstract： Aiming at the low accuracy of object detection methods based on unsupervised object detection, this paper proposes a foreground object detection method in unlabeled dataset. The basic idea is that correct feature clustering results can guide future object feature extraction, while the accurate foreground object features can improve the accuracy of feature clustering. The proposed method extracts local features from unlabeled images and then clusters features based on minimum feature distances. By matching pairwise images in the same cluster, feature weights are computed through feature correspondence. Finally, the updated feature weights are used to guide feature clustering in the next iteration. We simultaneously group similar images and detect foreground objects after iterations. The experimental results on Caltech256 and Google car side images demonstrate the effectiveness of our method. Furthermore, due to the present unsupervised object detection methods lacking of incremental learning ability, we propose an incremental implementation of our method. The experimental results show the incremental learning method can improve the computation speed greatly.

Key words: foreground object detection, unsupervised learning, feature clustering, incremental learning, feature extraction

aim at:针对
detection methods：n.检测方法

一种新的无监督前景目标检测方法 A New Unsupervised Foreground Object Detection Method相关推荐

【阅读论文】基于统计特征的无监督时间序列异常检测方法
文章目录摘要 1.介绍 2.相关工作 3.问题描述 4.方法 4.1 统计特征 4.2提取特征总结 4.3 学习阶段算法 4.4测试阶段算法 5.结果 5.1 YOB评价:单变量 5.3对合成数据集 ...
A Two-stage Unsupervised Approach for Low light Image Enhancement（一种两阶段无监督的微光图像增强方法）
主要参考文献及其收获 Unpaired image-to-image translation using cycle-consistent adversarial networks Deep reti ...
【3D目标检测】SMOKE: Single-Stage Monocular 3D Object Detection via Keypoint Estimation
目录前提概述细节网络结构特征提取网络检测头1:关键点检测检测头2:3D参数回归损失函数前提需要先了解下CenterNet [目标检测]Objects as Points 概述本文 ...
【2022 小目标检测综述】Towards Large-Scale Small Object Detection: Survey and Benchmarks
Towards Large-Scale Small Object Detection: Survey and Benchmarks 论文简介: 论文动机: 小目标检测方法回顾: 问题定义: 主要挑战: ...
对抗样本无法被重建！CMU提出通用的无监督对抗攻击检测方法
作者 | Ben Dickson 编译 | 琰琰机器学习在应用程序中的广泛使用,引起了人们对潜在安全威胁的关注.对抗性攻击( adversarial attacks)是一种常见且难以察觉的威胁手段, ...
一种关注于重要样本的目标检测方法！
↑↑↑关注后"星标"Datawhale 每日干货 & 每月组队学习,不错过 Datawhale学术作者:宋志龙,浙江工业大学,Datawhale成员在目标检测中训练模型 ...
【深度学习】一种关注于重要样本的目标检测方法！
作者:宋志龙,浙江工业大学,Datawhale成员在目标检测中训练模型时,样本间往往有差异性,不能被简单地同等对待.这次介绍的论文提出了一种重要样本的关注机制,在训练过程中帮助模型分辨哪些是重要的样 ...
显著目标检测论文(三)——Minimum Barrier Salient Object Detection at 80 FPS (2015)
这篇文章最大的亮点就是其实时性, 80 fps. 个人感觉论文的效果还是很惊艳的. 可以先看看论文的效果. 如 Figure 1 所示. 作者使用的机器配置如下: 3.2GHz x 2 CPU 12G ...
目标检测--Rich feature hierarchies for accurate object detection and semantic segmentation(CVPR 2014)
Rich feature hierarchies for accurate object detection and semantic segmentation 作者: Ross Girshick J ...

一种新的无监督前景目标检测方法 A New Unsupervised Foreground Object Detection Method

一种新的无监督前景目标检测方法 A New Unsupervised Foreground Object Detection Method相关推荐

最新文章

热门文章