gitkitten-优快云博客

原创【论文笔记】CVPR2020 Interpretable and Accurate Fine-grained Recognition via Region Grouping

Contributionproposed an interpretable deep model for fine-grained visual recognition:做细粒度分类，但同时output the segmentation of object parts and the identification of their contributions towards classification，增加了模型的可解释性为了确认object parts，使用a simple prior (pr.

2020-07-12 01:32:43 2071

原创【论文笔记】CVPR2020 A Multi-task Mean Teacher for Semi-supervised Shadow Detection

Multi-task Learning的工作：Contribuions提出了：利用更加丰富的unlabeled data利用multi-task learning来提供complementary information，让预测更准确来分别解决之前工作的问题：都需要 sufficient amount of annotated data，but annotated data are all captured in limited scenes.实验中作者发现之前的算法 ne-

2020-07-10 22:59:24 1144

原创【论文笔记】CVPR2020 Multi-scale Domain-adversarial Multiple-instance CNN for Cancer Subtype Classificatio

又一篇CVPR2020的histo image的文章，cancer subtype classificationContributioncancer subtype classification 面临三个问题：tumour and non-tumour regions are mixed, patch label is unavailablestaining conditions vary greatlywhen the magnification of image changes, di

2020-07-07 23:23:14 1135

原创【论文笔记】CVPR2020 Predicting Lymph Node Metastasis Using Histopathological Images Based on Multiple Ins

今年cvpr为什么这么多medical方向的文章，学习了新知识：multi-instance learningContribution作者解决了Predicting Lymph Node Metastasis Using Histopathological Images，由于whole slide images都太大了，所以提出了把大图分成patch来做multi-instance learning的想法。由于没有instance-level label，作者用self-supervised的

2020-07-06 12:03:27 753

原创【论文笔记】CVPR2020 Disentangling and Unifying Graph Convolutions for Skeleton-Based Action Recognition

使用GCN进行skeleton-based action recognitionContribution提出了两个设计：a disentangled multi-scale aggregation schemea unified spatial-temporal graph convolutional module (G3D)分别解决了两个问题：unbiased weight problem: edge weights will be biased towards closer node

2020-06-29 18:27:35 1534 4

原创【论文笔记】CVPR2020 Skeleton-Based Action Recognition with Shift Graph Convolutional Network

用图卷积做action recogntionContribution提出了shift-GCN（spatial shift graph operations & temporal shift graph operations）that exceeds the state-of-the-art methods with more than 10× less computational cost，主要解决了过去工作的两个问题heavy computational complexity of G

2020-06-29 00:30:47 1876

原创【论文笔记】CVPR2020 Rethinking Computer-aided Tuberculosis Diagnosis

在cvpr上少见的使用medical data的paperContributions收集了新的很大的TB dataset：Tuberculosis X-ray (TBX11K) dataset，包括：11200 X-ray ImagesImage-level annotation + TB area annotation using bounding boxesImage-level annotations include 4 classes: healthy, active TB, lat

2020-06-28 13:51:50 600

原创【论文笔记】CVPR2020 Exploring Self-attention for Image Recognition

CVPR 2020 的一篇自注意力机制Contributionsexplore variations of self-attention and assess their effectiveness for image recognition; 按两类self-attention进行探讨：pairwise self-attention & patchwise self-attention主要结论：MethodsPairwise Self-attention乘在bet

2020-06-27 17:21:52 1514

gitcat的博客