Bac. L: backbone for localization, does not exist for methods use a single network for classification and localization.
Top-1/Top-5 CLS: is correct if the Top-1/Top-5 predict categories contain the correct label.
GT-known Loc is correct when the intersection over union (IoU) between the ground-truth and the prediction is larger than 0.5 and does not consider whether the predicted category is correct.
Top-1/Top-5 Loc is correct when Top-1/Top-5 CLS and GT-Known LOC are both correct.
"-" indicates not exist. "?" indicates the corresponding item is not mentioned in the paper.
1.1. Top1/5 results on CUB-200-2011
Transformer
Method
Pub.
Bac.C
Bac.L
Top-1/5 Loc
GT-Known
Top-1/5 Cls
TS-CAM
2021ICCV
Deit-S
-
71.3/83.8
87.7
-/-
VGG
Method
Pub.
Bac.C
Bac.L
Top-1/5 Loc
GT-Known
Top-1/5 Cls
CREAM
2022CVPR
VGG16
-
70.4/85.7
91.0
PSOL
2020CVPR
VGG16
VGG16
66.3/84.1
-
-/-
GC-Net
2020ECCV
VGG16
VGG16
63.2/75.5
81.1
76.8/92.3
MEIL
2020CVPR
VGG16
-
57.5/-
73.8
74.8/-
DANet
2019ICCV
VGG16
-
52.5/62.0
67.7
75.4/92.3
CutMix
2019ICCV
VGG16
-
52.5/-
-
-
ADL
2019CVPR
VGG16
-
52.4/-
75.4
65.3/-
CAM
2016CVPR
VGG16
-
44.2/52.2
56.0
76.6/92.5
SPG
2018ECCV
VGG16
-
48.9/57.9
58.9
75.5/92.1
InceptionV3
Method
Pub.
Bac.C
Bac.L
Top-1/5 Loc
GT-Known
Top-1/5 Cls
CREAM
2022CVPR
InceptionV3
-
71.8/86.4
90.4
-/-
PSOL
2020CVPR
InceptionV3
InceptionV3
65.5/83.4
-
-/-
I2C
2020ECCV
InceptionV3
56.0/68.3
72.6
-/-
DANet
2019ICCV
InceptionV3
-
49.5/60.5
67.0
71.2/90.6
ADL
2019CVPR
InceptionV3
-
53.0/-
-
74.6/-
SPG
2018ECCV
InceptionV3
-
46.6/57.7
-
-
Others
Method
Pub.
Bac.C
Bac.L
Top-1/5 Loc
GT-Known
Top-1/5 Cls
ResNet50
DA-WSOL
2022CVPR
ResNet50
-
66.8/-
82.3
-/-
CutMix
2019ICCV
ResNet50
-
54.81/-
-
-/-
GoogleNet
CAM
2016CVPR
GoogleNet
-
41.1/50.7
-
73.8/91.5
1.2. Top1/5 results on ImageNet
Transformer
Method
Pub.
Bac.C
Bac.L
Top-1/5 Loc
GT-Known
Top-1/5 Cls
ViTOL
2022CVPRW
DeiT-B
-
58.6/-
72.5
77.1/-
TS-CAM
2021ICCV
Deit-S
-
53.4/64.3
67.6
-/-
VGG
Method
Pub.
Bac.C
Bac.L
Top-1/5 Loc
GT-Known
Top-1/5 Cls
CREAM
2022CVPR
VGG16
-
52.4/64.2
68.3
-/-
PSOL
2020CVPR
VGG16
VGG16
50.9/60.9
64.0
-/-
I2C
2020ECCV
VGG16
-
47.4/58.5
63.9
69.4/89.3
MEIL
2020CVPR
VGG16
-
46.8/-
-
70.3/-
ADL
2019CVPR
VGG16
-
44.9/-
-
69.5/-
CAM
2016CVPR
VGG16
-
42.8/54.9
59.0
68.8/88.6
InceptionV3
Method
Pub.
Bac.C
Bac.L
Top-1/5 Loc
GT-Known
Top-1/5 Cls
CREAM
2022CVPR
InceptionV3
-
56.1/66.2
69.0
-/-
PSOL
2020CVPR
InceptionV3
InceptionV3
54.8/63.3
65.2
-/-
I2C
2020ECCV
InceptionV3
-
53.1/64.1
68.5
73.3/91.6
GC-Net
2020ECCV
InceptionV3
InceptionV3
49.1/58.1
-
77.4/93.6
MEIL
2020CVPR
InceptionV3
-
49.5/-
-
73.3/-
ADL
2019CVPR
InceptionV3
-
48.7/-
-
72.8/-
SPG
2018ECCV
InceptionV3
-
48.6/60.0
64.7
CAM
2016CVPR
InceptionV3
-
46.3/58.2
62.7
73.3/91.8
Others
Method
Pub.
Bac.C
Bac.L
Top-1/5 Loc
GT-Known
Top-1/5 Cls
ResNet50
DA-WSOL
2022CVPR
ResNet50
-
54.9/-
70.2
-/-
CutMix
2019ICCV
ResNet50
-
47.25/-
-
78.6/94.1
GoogleNet
CAM
2016CVPR
GoogleNet
-
41.1/50.7
-
73.8/91.5
1.3. MaxBoxAccV2 results on CUB-200-2011
To do
1.4. MaxBoxAccV2 results on ImageNet
To do
2. Paper List
2022
CREAM:2022CVPR CREAM: Weakly Supervised Object Localization via Class RE-Activation Mapping
DA-WSOL:2022CVPR Weakly Supervised Object Localization as Domain Adaption
AlignMix:2022CVPR AlignMix: Improving representation by interpolating aligned features
ViTOL:2022CVPRW ViTOL: Vision Transformer for Weakly Supervised Object Localization
2022TPAMI Evaluation for Weakly Supervised Object Localization Protocol, Metrics, and Datasets
2022TNNLS Diverse Complementary Part Mining for Weakly Supervised Object Localization
2022PR Gradient-based refined class activation map for weakly supervised object localization
2022arxiv Learning Consistency from High-quality Pseudo-labels for Weakly Supervised Object Localization
I2C:2020ECCV Inter-Image Communication for Weakly Supervised Localization
2020ECCV Pairwise Similarity Knowledge Transfer for Weakly Supervised Object Localization
2020ICPR Dual-attention Guided Dropblock Module for Weakly Supervised Object Localization
2020arxiv Rethinking Localization Map Towards Accurate Object Perception with Self-Enhancement Maps
2019
ADL:2019CVPR Attention-based Dropout Layer for Weakly Supervised Object Localization
DANet:2019ICCV DANet: Divergent Activation for Weakly Supervised Object Localization
CutMix:2019ICCV CutMix: Regularization Strategy to Train Strong Classifiers with Localizable Features
2019ICLR Marginalized average attentional network for weakly-supervised learning
2019arxiv Dual-attention Focused Module for Weakly Supervised Object Localization
2019arxiv Weakly Supervised Localization Using Background Images
2019arxiv Weakly Supervised Object Localization with Inter-Intra Regulated CAMs
2018
ACoL:2018CVPR Adversarial Complementary Learning for Weakly Supervised Object Localization
SPG:2018ECCV Self-produced Guidance for Weakly-supervised Object Localization
2017
Grad-CAM:2017ICCV Grad-CAM: Visual Explanations from Deep Networks via Gradient-based Localization
HaS:2017ICCV Hide-and-Seek Forcing a Network to be Meticulous for Weakly-supervised Object and Action Localization
2016
CAM:2016CVPR Learning Deep Features for Discriminative Localization
3. Dataset
CUB-200-2011
@article{wah2011caltech,
title={The caltech-ucsd birds-200-2011 dataset},
author={Wah, Catherine and Branson, Steve and Welinder, Peter and Perona, Pietro and Belongie, Serge},
year={2011},
publisher={California Institute of Technology}
}
ImageNet
@inproceedings{deng2009imagenet,
title={Imagenet: A large-scale hierarchical image database},
author={Deng, Jia and Dong, Wei and Socher, Richard and Li, Li-Jia and Li, Kai and Fei-Fei, Li},
booktitle={2009 IEEE conference on computer vision and pattern recognition},
pages={248--255},
year={2009},
organization={Ieee}
}
About
Awesome weakly-supervised object localization, paper list, performance list