![](https://img-blog.csdnimg.cn/20201014180756923.png?x-oss-process=image/resize,m_fixed,h_64,w_64)
AI
燃烧橙
2019年南京邮电大学本科软件工程(嵌入式培养)毕业
寻实习
算法/JavaWeb方向
展开
-
Attributes-Guided and Pure-Visual Attention Alignment for Few-Shot Recognition笔记
方法:以半监督的方式。模块在有属性信息 和无属性信息的方式下,进行channel attention与spatial attention。数据集: CUB-200-2011,SUN Attribute Database类别:CUB:train:val:test = 2:1:1SUN: train:val:test = 580:65:72trained from scratchbackbone: Conv-4 and ResNet-12,...原创 2021-02-23 13:16:19 · 481 阅读 · 0 评论 -
BSNet: Bi-Similarity Network for Few-shot Fine-grained Image Classification笔记
方法:embedding module 提取特征,然后通过Bi-Similarity module双相似度模块来计算得分然后取均值数据集: FGVC-Aircraft, Stanford-Cars, Stanford-Dogs and CUB-200-2011类别:train:val:test = 2:1:1尺寸:84*84,不使用边框trained from scratchbackbone: Conv4, Conv64F结果...原创 2021-02-23 13:15:31 · 1224 阅读 · 0 评论