细粒度识别的挑战：类间差异大，类内差异小

常用的数据集：鸟，花，汽车……

不同的范式：

1. by localization-classification subnetworks

定位子网络，典型工作：S3N
ICCV2019:Selective Sparse Sampling for fine-grained image recognition
AAAI2020:Filtration and Distillation:Enhancing region attention for FGVC
AAAI2020:Graph-Propagation Based correlation learning for fgvc
CVPR2020:Weakly supervised fine-grained image classification vis gaussian mixture model oriented discriminative learning

2. by end-to-end feature encoding

ICCV2015:binlinear cnn
NIPS2019:learning deep bilinear transformation for fgvc
ICCV2019:cross-x learning for fgvc
AAAI2020:fine-grained recognition: accounting for subtle differences between similar classes
CVPR2020:fine-grained image-to-image transformation towards visual recognition
跨层，类内交互等

3.by leveraging attention mechanisms

CVPR2017:look closer to see better:recurrent attention convolutional neural network for fgvc
ICCV2019:learning a mixture of granularity-specific experts for fgvc
CVPR2020:attention convolutional binary neural tree for fgvc

4.by contrastive learning manners(新)

AAAI2020:learning attention pairwise interaction for fgvc
AAAI2020:channel interaction networks for fgvc

5.recognition with web data

AAAI2020:web-supervised network softlt update-drop training for fgvc
ACCV2020的比赛

6.recognition with limited data

TIP2019:piecewise classifier mappings:learning fine-grained learners for novel categories with few examples
IJCAI2020:multi-attention meta learning for few-shot fgvc
CVPR2020:reviseiting pose-normalization fir fgvc