细粒度分类:PMG论文笔记——Fine-Grained Visual Classification via Progressive Multi-Granularity Training of Jigsaw Patches
综述
论文题目:《Fine-Grained Visual Classification via Progressive Multi-Granularity Training of Jigsaw Patches》
会议时间:European Conference on Computer Vision 2020 (ECCV, 2020)
论文地址:https://arxiv.org/pdf/2003.03836.pdf
源码地址(PyTorch版本):https://github.com/PRIS-CV/PMG-Progressive-Multi-Granularity-Training
针对领域:细粒度图像分类(FGVC)
关键词:细粒度分类、渐进训练、拼图补丁
源码笔记:https://blog.csdn.net/qq_50001789/article/details/131360043?spm=1001.2014.3001.5501
主要思想
在以前的细粒度分类中,算法主要关注如何定位最有区分度的区域、互补区域以及多粒度区域,但是很少有算法关注如何将不同的粒度特征融合起来。