细粒度图像识别

Fine-grained Image Recognition

大四上半年一直在做细粒度图像识别的相关研究工作，做了一个学期，也摸索出一些方案，最终把精度做到了和ICCV 2019一样——87.7%（数据集：CUB-200-2011，其他数据集暂时还没有验证性能），后面想接着做下去，超过ICCV 2019，但是被导师制止了，说我创新点不足。。。无语。。。

接下来总结一下细粒度相关的研究进展：

1. Learning to Navigate for Fine-grained Classi cation—— ECCV 2018

datasets	top-1 Acc
CUB-200-2011	87.5%
FGVC Aircraft	91.4%
Stanford Cars	93.9%

在不同数据集上的精度，Backbone:ResNet 50

2.LEARNING TO FOCUS AND DISCRIMINATE FOR FINE-GRAINED CLASSIFICATION——ICIP 2019

datasets	top-1 Acc
CUB-200-2011	87.6%
FGVC Aircraft	92.1%
Stanford Cars	94.1%

在不同数据集上的精度，Backbone:ResNet 50

3.Cross-X Learning for Fine-Grained Visual Categorization——ICCV 2019

datasets	top-1 Acc
CUB-200-2011	87.7%
Stanford Dogs	88.9%
FGVC Aircraft	92.7%

在不同数据集上的精度，Backbone:ResNet 50

4.Destruction and Construction Learning for Fine-grained Image Recognition——CVPR 2019

datasets	top-1 Acc
CUB-200-2011	87.8%
Stanford Cars	94.5%
FGVC Aircraft	93.0%

在不同数据集上的精度，Backbone:ResNet 50

5.Looking for the Devil in the Details: Learning Trilinear Attention Sampling Network for Fine-grained Image Recognition——CVPR 2019

datasets	top-1 Acc
CUB-200-2011	87.9%
Stanford Cars	93.8%
iNaturalist 2017	68.2%

在不同数据集上的精度，Backbone:ResNet 50

6.Selective Sparse Sampling for Fine-grained Image Recognition——ICCV 2019

datasets	top-1 Acc
CUB-200-2011	88.5%
Stanford Cars	94.7%
Stanford Aircraft	92.8%

在不同数据集上的精度，Backbone:ResNet 50