Generalized Few-Shot Learning on ImageNet-LT

Metric: Top-1 Accuracy (higher is better)

LeaderboardDataset

Loading chart...

Results

Hide augmentations

Sort:

#	Model↕	Top-1 Accuracy▼	Augmentations	Paper	Date↕	Code
1	LIFT (ViT-L/14)	82.9	No	Long-Tail Learning with Foundation Model: Heavy ...	2023-09-18	Code
2	µ2Net+ (ViT-L/16)	82.5	No	A Continual Development Methodology for Large-sc...	2022-09-15	Code
3	MAM (ViT-B/16)	82.3	Yes	Improving Image Recognition by Retrieving from W...	2023-04-11	-
4	LIFT (ViT-B/16)	78.3	No	Long-Tail Learning with Foundation Model: Heavy ...	2023-09-18	Code
5	VL-LTR (ViT-B-16)	77.2	Yes	VL-LTR: Learning Class-wise Visual-Linguistic Re...	2021-11-26	Code
6	BALLAD(ResNet-50×16)	76.5	No	A Simple Long-Tailed Recognition Baseline via Vi...	2021-11-29	Code
7	BALLAD(ViT-B-16)	75.7	No	A Simple Long-Tailed Recognition Baseline via Vi...	2021-11-29	Code
8	BALLAD(ResNet-101)	70.5	No	A Simple Long-Tailed Recognition Baseline via Vi...	2021-11-29	Code
9	VL-LTR (ResNet-50)	70.1	Yes	VL-LTR: Learning Class-wise Visual-Linguistic Re...	2021-11-26	Code
10	BALLAD(ResNet-50)	67.2	No	A Simple Long-Tailed Recognition Baseline via Vi...	2021-11-29	Code
11	GPaCo (2-ResNeXt101-32x4d)	63.2	No	Generalized Parametric Contrastive Learning	2022-09-26	Code
12	MDCS (ResNeXt-50)	61.8	No	MDCS: More Diverse Experts with Consistency Self...	2023-08-19	Code
13	TADE(ResNeXt101-32x4d)	61.4	No	Self-Supervised Aggregation of Diverse Experts f...	2021-07-20	Code
14	ProCo (ResNet50)	60.2	No	Probabilistic Contrastive Learning for Long-Tail...	2024-03-11	Code
15	PaCo(ResNeXt101-32x4d)	60	No	Parametric Contrastive Learning	2021-07-26	Code
16	DeiT-LT	59.1	No	DeiT-LT Distillation Strikes Back for Vision Tra...	2024-04-03	Code
17	APA (SE-ResNext-50)	59.1	No	Adaptive Parametric Activation	2024-07-11	Code
18	TADE(ResNeXt-50)	58.8	No	Self-Supervised Aggregation of Diverse Experts f...	2021-07-20	Code
19	GML (ResNeXt-50)	58.8	No	Long-Tailed Recognition by Mutual Information Ma...	2023-05-02	Code
20	DirMixE(ResNeXt-50)	58.61	No	Harnessing Hierarchical Label Distribution Varia...	2024-05-13	Code
21	NCL(ResNeXt-50)	58.4	No	Nested Collaborative Learning for Long-Tailed Vi...	2022-03-29	Code
22	PaCo(ResNeXt-50)	58.2	No	Parametric Contrastive Learning	2021-07-26	Code
23	BS-CMO (ResNet-50)	58	No	The Majority Can Help The Minority: Context-rich...	2021-12-01	Code
24	ProCo (ResNeXt50)	58	No	Probabilistic Contrastive Learning for Long-Tail...	2024-03-11	Code
25	APA (SE-ResNet-50)	57.9	No	Adaptive Parametric Activation	2024-07-11	Code
26	CBD-ENS (ResNet-152)	57.7	No	Class-Balanced Distillation for Long-Tailed Visu...	2021-04-12	Code
27	ResLT(ResNeXt-50-3 experts)	57.6	No	ResLT: Residual Learning for Long-tailed Recogni...	2021-01-26	Code
28	BatchFormer(ResNet-50, PaCo)	57.4	No	BatchFormer: Learning to Explore Sample Relation...	2022-03-03	Code
29	NCL(ResNet-50)	57.4	No	Nested Collaborative Learning for Long-Tailed Vi...	2022-03-29	Code
30	Difficulty-Net (ResNet-50 using RandAugment, single model)	57.4	No	Difficulty-Net: Learning to Predict Difficulty f...	2022-09-07	Code
31	RIDE-DiVE	57.12	No	Distilling Virtual Examples for Long-tailed Reco...	2021-03-28	Code
32	BCL(ResNeXt-50)	57.1	No	Balanced Contrastive Learning for Long-Tailed Vi...	2022-07-19	Code
33	RIDE (ResNeXt-50)	56.4	No	Long-tailed Recognition by Routing Diverse Distr...	2020-10-05	Code
34	GLMC (ResNeXt-50)	56.3	No	Global and Local Mixture Consistency Cumulative ...	2023-05-15	Code
35	BatchFormer(ResNet-50, RIDE)	55.7	No	BatchFormer: Learning to Explore Sample Relation...	2022-03-03	Code
36	CBD-ENS (ResNet-50)	55.6	No	Class-Balanced Distillation for Long-Tailed Visu...	2021-04-12	Code
37	ResLT(ResNeXt101-32x4d)	55.1	No	ResLT: Residual Learning for Long-tailed Recogni...	2021-01-26	Code
38	OPeN (ResNeXt-50)	55.1	No	Pure Noise to the Rescue of Insufficient Data: I...	2021-12-16	Code
39	RIDE (ResNet-50)	54.9	No	Long-tailed Recognition by Routing Diverse Distr...	2020-10-05	Code
40	Difficulty-Net (ResNet-50 w/o using RandAugment, single model)	54	No	Difficulty-Net: Learning to Predict Difficulty f...	2022-09-07	Code
41	LTR-weight-balancing(ResNeXt-50)	53.9	No	Long-Tailed Recognition via Weight Balancing	2022-03-27	Code
42	DRO-LT	53.5	No	Distributional Robustness Loss for Long-tail Lea...	2021-04-07	-
43	DisAlign	53.4	No	Distribution Alignment: A Unified Framework for ...	2021-03-30	Code
44	DiVE	53.1	No	Distilling Virtual Examples for Long-tailed Reco...	2021-03-28	Code
45	LDAM + DRW + SAM	53.1	No	Escaping Saddle Points for Effective Generalizat...	2022-12-28	Code
46	LADE	53	No	Disentangling Label Distribution for Long-tailed...	2020-12-01	Code
47	ResLT(ResNeXt50)	52.9	No	ResLT: Residual Learning for Long-tailed Recogni...	2021-01-26	Code
48	MiSLAS	52.7	No	Improving Calibration for Long-Tailed Recognition	2021-04-01	Code
49	TSC(ResNet-50)	52.4	No	Targeted Supervised Contrastive Learning for Lon...	2021-11-27	Code
50	LDAM-DRS-RSG	51.8	No	RSG: A Simple but Effective Module for Learning ...	2021-06-18	Code
51	De-confound-TDE	51.8	No	Long-Tailed Classification by Keeping the Good a...	2020-09-28	Code
52	KCL	51.5	No	-	-	-
53	cRT + SSP	51.3	No	Rethinking the Value of Labels for Improving Cla...	2020-06-13	Code
54	Logit adjustment	51.3	No	Long-tail learning via logit adjustment	2020-07-14	Code
55	MetaSAug (ResNet-152)	50.03	No	MetaSAug: Meta Semantic Augmentation for Long-Ta...	2021-03-23	Code
56	MetaSAug with CE loss	47.39	No	MetaSAug: Meta Semantic Augmentation for Long-Ta...	2021-03-23	Code
57	BLT (DenseNet-121)	44.7	No	-	-	Code
58	Difficulty-Net (ResNet-10 w/o using RandAugment, single model	44.6	No	Difficulty-Net: Learning to Predict Difficulty f...	2022-09-07	Code
59	IEM	43.2	No	-	-	-
60	smDRAGON	42	No	From Generalized zero-shot learning to long-tail...	2020-04-05	Code
61	BALMS	41.8	No	Balanced Meta-Softmax for Long-Tailed Visual Rec...	2020-07-21	Code
62	CB LWS	41.4	No	Decoupling Representation and Classifier for Lon...	2019-10-21	Code
63	CBExperts	39.2	No	Long-Tailed Recognition Using Class-Balanced Exp...	2020-04-07	Code
64	LFME + OLTR	38.8	No	Learning From Multiple Experts: Self-paced Knowl...	2020-01-06	Code
65	CDB-loss (ResNet 10)	38.5	No	Class-Wise Difficulty-Balanced Loss for Solving ...	2020-10-05	Code
66	BLT (ResNet-10)	38	No	-	-	Code
67	OLTR	35.6	No	Large-Scale Long-Tailed Recognition in an Open W...	2019-04-10	Code
68	Online Feature Augmentation	35.3	No	Feature Space Augmentation for Long-Tailed Data	2020-08-09	-
69	Domain Adaptation	29.9	No	Rethinking Class-Balanced Methods for Long-Taile...	2020-03-24	Code

#1LIFT (ViT-L/14)SOTA
82.9
Top-1 Accuracy· 2023-09-18
Long-Tail Learning with Foundation Model: Heavy Fine-Tuning Hurts Code
#2µ2Net+ (ViT-L/16)SOTA
82.5
Top-1 Accuracy· 2022-09-15
A Continual Development Methodology for Large-scale Multitask Dynamic ML Systems Code
#3MAM (ViT-B/16)
82.3
Top-1 Accuracy· Augmentations· 2023-04-11
Improving Image Recognition by Retrieving from Web-Scale Image-Text Data
#4LIFT (ViT-B/16)
78.3
Top-1 Accuracy· 2023-09-18
Long-Tail Learning with Foundation Model: Heavy Fine-Tuning Hurts Code
#5VL-LTR (ViT-B-16)SOTA
77.2
Top-1 Accuracy· Augmentations· 2021-11-26
VL-LTR: Learning Class-wise Visual-Linguistic Representation for Long-Tailed Visual Recognition Code
#6BALLAD(ResNet-50×16)
76.5
Top-1 Accuracy· 2021-11-29
A Simple Long-Tailed Recognition Baseline via Vision-Language Model Code
#7BALLAD(ViT-B-16)
75.7
Top-1 Accuracy· 2021-11-29
A Simple Long-Tailed Recognition Baseline via Vision-Language Model Code
#8BALLAD(ResNet-101)
70.5
Top-1 Accuracy· 2021-11-29
A Simple Long-Tailed Recognition Baseline via Vision-Language Model Code
#9VL-LTR (ResNet-50)
70.1
Top-1 Accuracy· Augmentations· 2021-11-26
VL-LTR: Learning Class-wise Visual-Linguistic Representation for Long-Tailed Visual Recognition Code
#10BALLAD(ResNet-50)
67.2
Top-1 Accuracy· 2021-11-29
A Simple Long-Tailed Recognition Baseline via Vision-Language Model Code
#11GPaCo (2-ResNeXt101-32x4d)
63.2
Top-1 Accuracy· 2022-09-26
Generalized Parametric Contrastive Learning Code
#12MDCS (ResNeXt-50)
61.8
Top-1 Accuracy· 2023-08-19
MDCS: More Diverse Experts with Consistency Self-distillation for Long-tailed Recognition Code
#13TADE(ResNeXt101-32x4d)SOTA
61.4
Top-1 Accuracy· 2021-07-20
Self-Supervised Aggregation of Diverse Experts for Test-Agnostic Long-Tailed Recognition Code
#14ProCo (ResNet50)
60.2
Top-1 Accuracy· 2024-03-11
Probabilistic Contrastive Learning for Long-Tailed Visual Recognition Code
#15PaCo(ResNeXt101-32x4d)
60
Top-1 Accuracy· 2021-07-26
Parametric Contrastive Learning Code
#16DeiT-LT
59.1
Top-1 Accuracy· 2024-04-03
DeiT-LT Distillation Strikes Back for Vision Transformer Training on Long-Tailed Datasets Code
#17APA (SE-ResNext-50)
59.1
Top-1 Accuracy· 2024-07-11
Adaptive Parametric Activation Code
#18TADE(ResNeXt-50)
58.8
Top-1 Accuracy· 2021-07-20
Self-Supervised Aggregation of Diverse Experts for Test-Agnostic Long-Tailed Recognition Code
#19GML (ResNeXt-50)
58.8
Top-1 Accuracy· 2023-05-02
Long-Tailed Recognition by Mutual Information Maximization between Latent Features and Ground-Truth Labels Code
#20DirMixE(ResNeXt-50)
58.61
Top-1 Accuracy· 2024-05-13
Harnessing Hierarchical Label Distribution Variations in Test Agnostic Long-tail Recognition Code
#21NCL(ResNeXt-50)
58.4
Top-1 Accuracy· 2022-03-29
Nested Collaborative Learning for Long-Tailed Visual Recognition Code
#22PaCo(ResNeXt-50)
58.2
Top-1 Accuracy· 2021-07-26
Parametric Contrastive Learning Code
#23BS-CMO (ResNet-50)
58
Top-1 Accuracy· 2021-12-01
The Majority Can Help The Minority: Context-rich Minority Oversampling for Long-tailed Classification Code
#24ProCo (ResNeXt50)
58
Top-1 Accuracy· 2024-03-11
Probabilistic Contrastive Learning for Long-Tailed Visual Recognition Code
#25APA (SE-ResNet-50)
57.9
Top-1 Accuracy· 2024-07-11
Adaptive Parametric Activation Code
#26CBD-ENS (ResNet-152)SOTA
57.7
Top-1 Accuracy· 2021-04-12
Class-Balanced Distillation for Long-Tailed Visual Recognition Code
#27ResLT(ResNeXt-50-3 experts)SOTA
57.6
Top-1 Accuracy· 2021-01-26
ResLT: Residual Learning for Long-tailed Recognition Code
#28BatchFormer(ResNet-50, PaCo)
57.4
Top-1 Accuracy· 2022-03-03
BatchFormer: Learning to Explore Sample Relationships for Robust Representation Learning Code
#29NCL(ResNet-50)
57.4
Top-1 Accuracy· 2022-03-29
Nested Collaborative Learning for Long-Tailed Visual Recognition Code
#30Difficulty-Net (ResNet-50 using RandAugment, single model)
57.4
Top-1 Accuracy· 2022-09-07
Difficulty-Net: Learning to Predict Difficulty for Long-Tailed Recognition Code
#31RIDE-DiVE
57.12
Top-1 Accuracy· 2021-03-28
Distilling Virtual Examples for Long-tailed Recognition Code
#32BCL(ResNeXt-50)
57.1
Top-1 Accuracy· 2022-07-19
Balanced Contrastive Learning for Long-Tailed Visual Recognition Code
#33RIDE (ResNeXt-50)SOTA
56.4
Top-1 Accuracy· 2020-10-05
Long-tailed Recognition by Routing Diverse Distribution-Aware Experts Code
#34GLMC (ResNeXt-50)
56.3
Top-1 Accuracy· 2023-05-15
Global and Local Mixture Consistency Cumulative Learning for Long-tailed Visual Recognitions Code
#35BatchFormer(ResNet-50, RIDE)
55.7
Top-1 Accuracy· 2022-03-03
BatchFormer: Learning to Explore Sample Relationships for Robust Representation Learning Code
#36CBD-ENS (ResNet-50)
55.6
Top-1 Accuracy· 2021-04-12
Class-Balanced Distillation for Long-Tailed Visual Recognition Code
#37ResLT(ResNeXt101-32x4d)
55.1
Top-1 Accuracy· 2021-01-26
ResLT: Residual Learning for Long-tailed Recognition Code
#38OPeN (ResNeXt-50)
55.1
Top-1 Accuracy· 2021-12-16
Pure Noise to the Rescue of Insufficient Data: Improving Imbalanced Classification by Training on Random Noise Images Code
#39RIDE (ResNet-50)
54.9
Top-1 Accuracy· 2020-10-05
Long-tailed Recognition by Routing Diverse Distribution-Aware Experts Code
#40Difficulty-Net (ResNet-50 w/o using RandAugment, single model)
54
Top-1 Accuracy· 2022-09-07
Difficulty-Net: Learning to Predict Difficulty for Long-Tailed Recognition Code
#41LTR-weight-balancing(ResNeXt-50)
53.9
Top-1 Accuracy· 2022-03-27
Long-Tailed Recognition via Weight Balancing Code
#42DRO-LT
53.5
Top-1 Accuracy· 2021-04-07
Distributional Robustness Loss for Long-tail Learning
#43DisAlign
53.4
Top-1 Accuracy· 2021-03-30
Distribution Alignment: A Unified Framework for Long-tail Visual Recognition Code
#44DiVE
53.1
Top-1 Accuracy· 2021-03-28
Distilling Virtual Examples for Long-tailed Recognition Code
#45LDAM + DRW + SAM
53.1
Top-1 Accuracy· 2022-12-28
Escaping Saddle Points for Effective Generalization on Class-Imbalanced Data Code
#46LADE
53
Top-1 Accuracy· 2020-12-01
Disentangling Label Distribution for Long-tailed Visual Recognition Code
#47ResLT(ResNeXt50)
52.9
Top-1 Accuracy· 2021-01-26
ResLT: Residual Learning for Long-tailed Recognition Code
#48MiSLAS
52.7
Top-1 Accuracy· 2021-04-01
Improving Calibration for Long-Tailed Recognition Code
#49TSC(ResNet-50)
52.4
Top-1 Accuracy· 2021-11-27
Targeted Supervised Contrastive Learning for Long-Tailed Recognition Code
#50LDAM-DRS-RSG
51.8
Top-1 Accuracy· 2021-06-18
RSG: A Simple but Effective Module for Learning Imbalanced Datasets Code
#51De-confound-TDESOTA
51.8
Top-1 Accuracy· 2020-09-28
Long-Tailed Classification by Keeping the Good and Removing the Bad Momentum Causal Effect Code
#52KCL
51.5
Top-1 Accuracy
No paper
#53cRT + SSPSOTA
51.3
Top-1 Accuracy· 2020-06-13
Rethinking the Value of Labels for Improving Class-Imbalanced Learning Code
#54Logit adjustment
51.3
Top-1 Accuracy· 2020-07-14
Long-tail learning via logit adjustment Code
#55MetaSAug (ResNet-152)
50.03
Top-1 Accuracy· 2021-03-23
MetaSAug: Meta Semantic Augmentation for Long-Tailed Visual Recognition Code
#56MetaSAug with CE loss
47.39
Top-1 Accuracy· 2021-03-23
MetaSAug: Meta Semantic Augmentation for Long-Tailed Visual Recognition Code
#57BLT (DenseNet-121)
44.7
Top-1 Accuracy
No paperCode
#58Difficulty-Net (ResNet-10 w/o using RandAugment, single model
44.6
Top-1 Accuracy· 2022-09-07
Difficulty-Net: Learning to Predict Difficulty for Long-Tailed Recognition Code
#59IEM
43.2
Top-1 Accuracy
No paper
#60smDRAGONSOTA
42
Top-1 Accuracy· 2020-04-05
From Generalized zero-shot learning to long-tail with class descriptors Code
#61BALMS
41.8
Top-1 Accuracy· 2020-07-21
Balanced Meta-Softmax for Long-Tailed Visual Recognition Code
#62CB LWSSOTA
41.4
Top-1 Accuracy· 2019-10-21
Decoupling Representation and Classifier for Long-Tailed Recognition Code
#63CBExperts
39.2
Top-1 Accuracy· 2020-04-07
Long-Tailed Recognition Using Class-Balanced Experts Code
#64LFME + OLTR
38.8
Top-1 Accuracy· 2020-01-06
Learning From Multiple Experts: Self-paced Knowledge Distillation for Long-tailed Classification Code
#65CDB-loss (ResNet 10)
38.5
Top-1 Accuracy· 2020-10-05
Class-Wise Difficulty-Balanced Loss for Solving Class-Imbalance Code
#66BLT (ResNet-10)
38
Top-1 Accuracy
No paperCode
#67OLTRSOTA
35.6
Top-1 Accuracy· 2019-04-10
Large-Scale Long-Tailed Recognition in an Open World Code
#68Online Feature Augmentation
35.3
Top-1 Accuracy· 2020-08-09
Feature Space Augmentation for Long-Tailed Data
#69Domain Adaptation
29.9
Top-1 Accuracy· 2020-03-24
Rethinking Class-Balanced Methods for Long-Tailed Visual Recognition from a Domain Adaptation Perspective Code