Fine Grained Image Classification
细粒度图像分类是计算机视觉中的一个任务,旨在将图像划分到更具体的子类别中。该任务要求模型能够识别和区分同一大类别内的细微视觉差异和模式,具有较高的挑战性。其应用价值在于提升图像识别的精度和细致度,适用于生物物种鉴定、商品分类等场景。
10 Monkey Species
Bird-225
WideResNet-101 (Spinal FC)
Birdsnap
EffNet-L2 (SAM)
Bottles
BoxCars116K
Caltech-101
CarFlag-1532
CarFlag-563
ResNet101-swp
CompCars
Resnet50 + PMAL
Con-Text
PHOC descriptor + Fisher Vector Encoding
CUB-200-2011
CAP
CUB-200-2011
MetaFormer
(MetaFormer-2,384)
DIB-10K
MetaFGNet
EMNIST-Digits
VGG-5
EMNIST-Letters
VGG-5
FGVC Aircraft
I2-HOFI
FGVC-Aircraft
EnGraf-Net101 (G=4, H=1)
Food-101
CAP
FoodX-251
CSWin-L
Fruits-360
VGG-19bn
Herbarium 2021 Half–Earth
Herbarium 2022
Conviformer-B
Imbalanced CUB-200-2011
PC-Softmax
iNaturalist
TASN
Kuzushiji-MNIST
MNIST
Vanilla FC layer only
NABirds
HERBS
Oxford 102 Flowers
IELT
Oxford-IIIT Pet Dataset
Oxford-IIIT Pets
µ2Net+ (ViT-L/16)
QMNIST
VGG-5
SOP
Assemble-ResNet-FGVC-50
Stanford Cars
Stanford Dogs
MP
STL-10
Pre trained wide-resnet-101
SUN397
µ2Net (ViT-L/16)