Image Retrieval on iNaturalist

Metric: R@32 (higher is better)

LeaderboardDataset

Loading chart...

Results

Submit a result

Sort:

#	Model↕	R@32▼	Extra Data	Paper	Date↕	Code
1	Recall@k Surrogate loss (ViT-B/16)	97.2	No	Recall@k Surrogate Loss with Large Batches and S...	2021-08-25	Code
2	ROADMAP (DeiT-S)	95.2	No	Robust and Decomposable Average Precision for Im...	2021-10-01	Code
3	Recall@k Surrogate loss (ResNet-50)	94.3	No	Recall@k Surrogate Loss with Large Batches and S...	2021-08-25	Code
4	ROADMAP (ResNet-50)	93.9	No	Robust and Decomposable Average Precision for Im...	2021-10-01	Code
5	Smooth-AP	93.1	No	Smooth-AP: Smoothing the Path Towards Large-Scal...	2020-07-23	Code
6	PNP Loss	92.6	No	Rethinking the Optimization of Average Precision...	2021-02-09	Code

#1Recall@k Surrogate loss (ViT-B/16)SOTA
97.2
R@32· 2021-08-25
Recall@k Surrogate Loss with Large Batches and Similarity Mixup Code
#2ROADMAP (DeiT-S)
95.2
R@32· 2021-10-01
Robust and Decomposable Average Precision for Image Retrieval Code
#3Recall@k Surrogate loss (ResNet-50)
94.3
R@32· 2021-08-25
Recall@k Surrogate Loss with Large Batches and Similarity Mixup Code
#4ROADMAP (ResNet-50)
93.9
R@32· 2021-10-01
Robust and Decomposable Average Precision for Image Retrieval Code
#5Smooth-APSOTA
93.1
R@32· 2020-07-23
Smooth-AP: Smoothing the Path Towards Large-Scale Image Retrieval Code
#6PNP Loss
92.6
R@32· 2021-02-09
Rethinking the Optimization of Average Precision: Only Penalizing Negative Instances before Positive Ones is Enough Code