Weakly supervised learning approaches for object segmentation

最新推荐文章于 2024-12-11 17:37:56 发布

乐兮山南水北

最新推荐文章于 2024-12-11 17:37:56 发布

阅读量416

点赞数 1

分类专栏：理论学习

本文链接：https://blog.youkuaiyun.com/u012494820/article/details/87889408

版权

理论学习专栏收录该内容

5 篇文章

订阅专栏

本文探讨了在弱监督学习环境下，通过使用正则化损失函数改进伪全标注掩膜的质量，以增强CNN在图像分割任务中的表现。当前的弱监督学习方法主要依赖于这些伪标签，但其准确性直接影响到模型的学习效果。

摘要生成于 C知道，由 DeepSeek-R1 满血版支持，前往体验 >

On Regularized Losses for Weakly-supervised CNN Segmentation Meng

Motivation

Currently, the dominated weakly supervised learning approaches learn from pseudo fully labeled masks.

确定要放弃本次机会？

福利倒计时

: :

立减 ¥

普通VIP年卡可用

立即使用

乐兮山南水北

关注关注

1
点赞
踩
0

收藏

觉得还不错? 一键收藏
0
评论
分享

复制链接

分享到 QQ

分享到新浪微博

扫一扫
举报

举报

专栏目录

How to do Semantic Segmentation using Deep learning(如何使用深度学习进行语义分割)

weixin_41697507的博客

04-14

806

This article is a comprehensive overview including a step-by-step guide to implement a deep learning image segmentation model. update: We just launched a new product: [Nanonets Object Detection APIs ]...

语义分割进阶之路之回首2015cvpr（三）

lirunqi3的博客

08-05

532

Weakly Supervised Semantic Segmentation for Social Images[full paper][ext. abstract] Wei Zhang, Sheng Zeng, Dequan Wang, Xiangyang Xue 因为这些文章并没有许多人有深入研究，所以就没有什么直接参考，所以我打算以边翻译边总结的方式学习这个算法。 abst...

参与评论您还未登录，请先登录后发表或查看评论

Weakly Supervised Deep Detection Networks 学习笔记

qq_35408241的博客

10-24

2359

Weakly Supervised Deep Detection Networks 详细解读论文大致意思是通过图像级的标注信息训练网络达到目标检测的目的，文中采用的方法如下图所示： abstract ：提出一种弱监督的深度检测架构，通过修改网络，只利用图像级别的标注，同时执行...

CAMEL: A Weakly Supervised Learning Framework for Histopathology Image Segmentation

mjiansun的专栏

08-01

1702

论文：https://arxiv.org/abs/1908.10555 参考网页。今天是20190801，暂时没能搜到该篇文章。但根据参考网页进行了学习，然后将自己的理解整理成了博客。简介在本研究中，我们提出了弱监督学习框架CAMEL，整套框架仅使用图像级别的标注即可完成像素级识别模型的搭建。CAMEL包括两个步骤：监督信息增强（label enrichment）和有监督模型训练（seg...

浅谈弱监督学习（Weakly Supervised Learning）

idol24的博客

04-11

3485

来源：知乎—Lori地址：https://zhuanlan.zhihu.com/p/81404885之前看了周志华老师的《A Brief Introduction to Weakly Supervised Learning》，觉得大受启发，一直想动笔总结一下，今天终于来填坑了。机器学习在很多任务中取得了很大的成功。众所周知，在机器学习领域，学习任务可大致划分为两类，一种是...

【配准】弱监督(Weakly-Supervised)系列配准论文阅读

m0_37935211的博客

11-01

4203

弱监督系列论文汇总表序号发表情况要点 1 ISBI2018 弱监督配准框架 2 MIA2018 多尺度Dice，新的网络，大量的实验 3 TMI2018 ...

弱监督检测初识——Weakly Supervised Deep Detection Networks解读

Cai Yichao的博客

11-26

3867

文章目录1 WSDDN模型1.1 预训练CNN1.2 CNN特征描述1.2.1 区域推荐算法1.2.2 Spatial partial pooling1.2.3 推荐区域的特征描述1.3 分类和检测分支2 模型训练2.1 损失函数2.2 检测框规则化3 局限与未来方向4 Reference 弱监督检测是一个很有前景的研究方向——仅通过样本的类别标签就可以实现目标检测任务（乍一看有点类似之前的Classification and localization任务，但是要难很多）。目前，弱监督检测的精度还处于比较低

CVPR 2020——Weakly Supervised Semantic Point Cloud Segmentation: Towards 10× Fewer Labels

Dujing2019的博客

05-09

1610

Weakly Supervised Semantic Point Cloud Segmentation: Towards 10× Fewer LabelsAbstract（一）Introduction（二） Related Work（三）Methodology3.1 Point Cloud Encoder Network3.2. Incomplete Supervision Branch3.3. Inexact Supervision Branch3.4. Siamese Self-Supervision3

拟投顶会的论文详细提纲“Weakly Supervised 3D Human Mesh Recovery from Depth Images using Sparse Annotations”

最新发布

结构化论文学习和写作

12-11

582

Title: Weakly Supervised 3D Human Mesh Recovery from Depth Images using Sparse AnnotationsAbstract: This paper addresses the challenge of 3D human mesh recovery from depth images in scenarios where full 3D ground truth annotations are scarce. We propose a

【AI视野·今日CV 计算机视觉论文速览第224期】Tue, 22 Jun 2021

TomRen

06-24

2528

AI视野·今日CS.CV 计算机视觉论文速览 Tue, 22 Jun 2021 (showing first 100 of 122 entries) Totally 100 papers ????上期速览✈更多精彩请移步主页 Daily Computer Vision Papers Towards Long-Form Video Understanding Authors Chao Yuan Wu, Philipp Kr henb hl我们的世界提供了永无止境的视觉刺激流，但今天的Vision

《Sentiment Analysis Based on Deep Learning Approaches》阅读笔记

t_zht的博客

12-30

504

2018 IEEE 参考意义不大【Lexicon is further divided into dictionary based and corpus based approach which further uses semantic and statistical methods to get the sentiment polarity of the text used in the...

【弱监督显著目标检测论文】Weakly-Supervised Salient Object Detection via Scribble Annotations

ooooollll111的博客

10-18

2646

与费力的像素级密集标注相比，用涂鸦标注数据要容易得多，标注一张图片只需花费1∼2秒。然而，使用涂鸦标签来学习显著目标检测还没有被探索过。在本文中，我们提出了一个弱监督的显著目标检测模型，从这种标注中学习显著性。在此过程中，我们首先重新用涂鸦标注了现有的大规模显著目标检测数据集，记为S-DUTS。由于物体的结构和细节信息不被涂鸦所识别，直接用涂鸦标签进行训练会导致显著图的边界定位不佳。为了缓解这个问题，我们提出了一个辅助的边缘检测任务来明确定位物体的边缘，以及一个门控结构感知损失来对要恢复的结构范围进行约束。

【论文阅读】【TCSVT2023】基于混合标签的显著目标检测弱监督学习框架

qq_52157805的博客

11-01

1086

目前全监督显著目标检测（SOD）方法已经取得了很大的进展，但这种方法往往依赖于大量的像素级标注，耗时且费力。本文研究了一种新的混合标签下的弱监督SOD任务，该任务的监督标签包括传统无监督方法生成的大量粗标签和少量真实标签。为了解决该任务中的标签噪声和数量不平衡问题，我们设计了一个新的管道框架，其中包含三种复杂的训练策略。在模型框架方面，我们将任务解耦为标签细化子任务和显著目标检测子任务，它们相互配合，交替训练。

弱监督/无监督显著目标检测(SOD)部分经典论文摘要

xiongxyowo的博客

10-10

3100

(CVPR’17) Learning to Detect Salient Objects with Image-level Supervision Deep Neural Networks (DNNs) have substantially improved the state-of-the-art in salient object detection. However, training DNNs requires costly pixel-level annotations. In this pape

【论文笔记】Weakly Supervised Learning Based on Coupled Convolutional Neural Networks for Aircraft Detecti

jiusake的博客

04-15

865

Weakly Supervised Learning Based on Coupled Convolutional Neural Networks for Aircraft Detection基于耦合卷积神经网络的弱监督学习用于飞机检测Fan Zhang, Bo Du, Senior Member, IEEE, Liangpei Zhang, Senior Member, IEEE, and Mi...

ECCV 2018 完整论文集 -- List & 下载链接

热门推荐

TomRen

09-02

1万+

下文列表为ECCV2018官网得到了今年接收论文列表，共779篇：持续更新下载链接 Oral: Convolutional Networks with Adaptive Computation Graphs Progressive Neural Architecture Search Diverse Image-to-Image Translation via Dise...

【论文总结】weakly- and semi-supervised learning of a DCNN for semantic Image Segmentation

u011582187的专栏

05-13

5699

一、概述这篇文章研究了如何从弱注释的训练数据（如边界框或图像级标签）或少量强标记图像和许多弱标记图像的组合中学习DCNN用于语义图像分割的问题，在弱超监督和半监督条件下提出了期望最大化（EM）方法。代码：https://bitbucket.org/deeplab/deeplab-public二、研究内容文章将焦点放在用弱标签训练调参上，所以训练阶段没使用CRF，只在测试阶段使用。编码：m表示像素，...

图像质量评价(IQA)论文笔记：Blind Image Quality Assessment via Deep Learning

VisualCortex

04-12

8629

IQA论文笔记：Blind Image Quality Assessment via Deep Learning Weilong Hou et al IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS, VOL. 26, NO. 6, JUNE 2015 INTRO 对于IQA的label，也就是通常用来做回归的score...

ICCV 2019 | 旷视研究院提出基于互向导的半监督皮肤检测新方法

旷视开发者技术社区

08-23

1131

两年一度的国际计算机视觉大会 ICCV 2019 ( IEEE International Conference on Computer Vision) 将于当地时间 1...

Regional Semantic Contrast and Aggregation for Weakly Supervised Semantic Segmentation

04-05

Weakly supervised semantic segmentation is a challenging task where only image-level labels are available for training. In this paper, we propose a novel approach that utilizes regional semantic contrast and aggregation to improve the performance of weakly supervised semantic segmentation. First, we introduce a regional semantic contrast loss that encourages the model to focus on regions with high semantic contrast, i.e., regions where the difference between the foreground and background classes is high. Specifically, we compute the semantic contrast map by subtracting the predicted probability of the background class from the predicted probability of the foreground class, and then apply a Gaussian filter to smooth the map. Second, we propose a regional aggregation module that aggregates the features of each region to obtain a more representative feature representation. The module consists of a region proposal network that generates candidate regions based on the semantic contrast map, and a region-based pooling layer that pools the features within each region. Finally, we combine the regional semantic contrast loss and the regional aggregation module with a standard cross-entropy loss to train the model. Experimental results on the PASCAL VOC 2012 dataset demonstrate that our approach achieves state-of-the-art performance among weakly supervised methods, and even outperforms some fully supervised methods.