
细粒度分类
ama_momo
这个作者很懒,什么都没留下…
展开
-
Focus Longer to See Better: Recursively Refined Attention for Fine-Grained Image Classification
Focus Longer to See Better: Recursively Refined Attention for Fine-Grained Image Classificationcode:https://github.com/TAMU-VITA/Focus-Longer-to-See-Betterpaper:https://arxiv.org/abs/2005.10979Abstract类间的边缘视觉差异(the marginal visual difference)使得细粒度分类很难。原创 2020-07-15 18:06:24 · 484 阅读 · 1 评论 -
Regularized Pooling
Abstract在CNNs中pooling的作用:dimensionality reduction and deformation compensation.但是存在问题:its excessive flexibility risks canceling the essential spatial differences between classes.(过度灵活性可能会抵消类之间的本质空间差异)提出:regularized pooling,提高了识别精度,而且加快了学习的收敛速度。1 Introdu原创 2020-06-08 10:25:56 · 262 阅读 · 1 评论 -
Fine-Grained Visual Classification via Progressive Multi-Granularity Training of Jigsaw Patches
基于渐进式多粒度拼图训练的细粒度视觉分类code:https://github.com/PRIS-CV/PMG-Progressive-Multi-Granularity-Trainingpaper:https://arxiv.org/abs/2003.03836Abstract细粒度分类比传统分类难目前解决方法主要关注:locate the most discriminative parts, more complementary parts, and parts of various gran原创 2020-06-08 10:08:16 · 645 阅读 · 3 评论 -
Learning to Navigate for Fine-grained Classification
ECCV 2018 北京大学Abstract找出完全表征对象的细微特征并不简单(细粒度分类的挑战性)文章提出新颖的自监督(self-supervision)机制,无需bbox和part annotations,即可有效定位信息区域。模型:NTS-Net( Navigator-Teacher-Scrutinizer Network)== Navigator agent,Teacher agent和Scrutinizer agent组成考虑到informativeness of the regions原创 2020-06-04 12:19:36 · 723 阅读 · 2 评论 -
双线性卷积神经网络模型(Bilinear CNN)
ICCV 2015参考资料摘要双线性CNN模型:包含两个特征提取器,其输出经过外积相乘,池化后获得image descriptorfeature fusion的方式有很多,所以在两个数据集上进行了三种方式的feature fusion结果比较。很多广泛使用的texture representation可以被表示为两个设计合理的特征的outer product。2018DataFold-1Fold-2Fold-3Fold-4Fold-5avg外积0.90309原创 2020-06-04 12:04:08 · 15839 阅读 · 8 评论 -
Fine-Grained Visual Classification via Progressive Multi-Granularity Training of Jigsaw Patches
一. Fine-Grained Visual Classification via Progressive Multi-Granularity Training of Jigsaw Patches基于渐进式多粒度拼图训练的细粒度视觉分类code:https://github.com/PRIS-CV/PMG-Progressive-Multi-Granularity-Trainingpaper:https://arxiv.org/abs/2003.03836Abstract细粒度分类比传统分类难目前原创 2020-06-03 18:40:11 · 2037 阅读 · 12 评论 -
细粒度分类:HBP
Hierarchical Bilinear Pooling for Fine-Grained Visual RecognitionECCV 2018 华中科技大学论文代码1. Abstract在细粒度图像分类中,双线性池化(bilinear pooling)的模型已经被证明是有效的,然而,先前的大多方法忽略了这样一个事实:层间部分特征交互和细粒度特征学习是相互关联的并且可以相互加强。根据...原创 2020-04-20 16:09:50 · 1994 阅读 · 1 评论