论文阅读 [TPAMI-2022] Leveraging Instance-, Image- and Dataset-Level Information for Weakly Supervised I

论文阅读 [TPAMI-2022] Leveraging Instance-, Image- and Dataset-Level Information for Weakly Supervised Instance Segmentation

论文搜索(studyai.com)

搜索论文: Leveraging Instance-, Image- and Dataset-Level Information for Weakly Supervised Instance Segmentation

搜索论文: http://www.studyai.com/search/whole-site/?q=Leveraging+Instance-,+Image-+and+Dataset-Level+Information+for+Weakly+Supervised+Instance+Segmentation

关键字(Keywords)

Semantics; Proposals; Image segmentation; Training; Probability distribution; Feature extraction; Noise measurement; Weakly supervised learning; instance segmentation; semantic segmentation; multiple instance learning; multi-way cut

机器学习; 机器视觉; 自然语言处理

监督学习; 检测分割; 知识图谱; 弱监督学习; 多实例学习; 语义分析; 语义特征

摘要(Abstract)

Weakly supervised semantic instance segmentation with only image-level supervision, instead of relying on expensive pixel-wise masks or bounding box annotations, is an important problem to alleviate the data-hungry nature of deep learning.

弱监督语义实例分割是一个重要问题，它只需要图像级的监督，而不是依赖昂贵的像素级掩码或边界框注释，以缓解深度学习的数据饥饿性。.

In this article, we tackle this challenging problem by aggregating the image-level information of all training images into a large knowledge graph and exploiting semantic relationships from this graph.

在本文中，我们通过将所有训练图像的图像级信息聚合到一个大型知识图中，并利用该图中的语义关系来解决这个具有挑战性的问题。.

Specifically, our effort starts with some generic segment-based object proposals (SOP) without category priors.

具体来说，我们的工作从一些通用的基于段的对象建议（SOP）开始，没有类别优先级。.

We propose a multiple instance learning (MIL) framework, which can be trained in an end-to-end manner using training images with image-level labels.

我们提出了一个多实例学习（MIL）框架，该框架可以使用带有图像级标签的训练图像以端到端的方式进行训练。.

For each proposal, this MIL framework can simultaneously compute probability distributions and category-aware semantic features, with which we can formulate a large undirected graph.

对于每一个方案，这个MIL框架都可以同时计算概率分布和类别感知语义特征，利用这些特征我们可以形成一个大的无向图。.

The category of background is also included in this graph to remove the massive noisy object proposals.

该图中还包括背景类别，以去除大量噪声对象。.

An optimal multi-way cut of this graph can thus assign a reliable category label to each proposal.

因此，该图的最佳多路切割可以为每个提案分配一个可靠的类别标签。.

The denoised SOP with assigned category labels can be viewed as pseudo instance segmentation of training images, which are used to train fully supervised models.

将带指定类别标签的去噪SOP视为训练图像的伪实例分割，用于训练完全监督模型。.

The proposed approach achieves state-of-the-art performance for both weakly supervised instance segmentation and semantic segmentation.

该方法在弱监督的实例分割和语义分割方面都取得了最新的性能。.

The code is available at https://github.com/yun-liu/LIID…

该代码可在https://github.com/yun-liu/LIID…

作者(Authors)

[‘Yun Liu’, ‘Yu-Huan Wu’, ‘Peisong Wen’, ‘Yujun Shi’, ‘Yu Qiu’, ‘Ming-Ming Cheng’]