Image Retrieval on iNaturalist

Metric: R@16 (higher is better)

LeaderboardDataset

Loading chart...

Results

Submit a result

Sort:

#	Model↕	R@16▼	Extra Data	Paper	Date↕	Code
1	Recall@k Surrogate loss (ViT-B/16)	95.9	No	Recall@k Surrogate Loss with Large Batches and S...	2021-08-25	Code
2	ROADMAP (DeiT-S)	93.1	No	Robust and Decomposable Average Precision for Im...	2021-10-01	Code
3	Recall@k Surrogate loss (ResNet-50)	91.9	No	Recall@k Surrogate Loss with Large Batches and S...	2021-08-25	Code
4	ROADMAP (ResNet-50)	91.3	No	Robust and Decomposable Average Precision for Im...	2021-10-01	Code
5	Smooth-AP	90.3	No	Smooth-AP: Smoothing the Path Towards Large-Scal...	2020-07-23	Code
6	PNP Loss	89.7	No	Rethinking the Optimization of Average Precision...	2021-02-09	Code

#1Recall@k Surrogate loss (ViT-B/16)SOTA
95.9
R@16· 2021-08-25
Recall@k Surrogate Loss with Large Batches and Similarity Mixup Code
#2ROADMAP (DeiT-S)
93.1
R@16· 2021-10-01
Robust and Decomposable Average Precision for Image Retrieval Code
#3Recall@k Surrogate loss (ResNet-50)
91.9
R@16· 2021-08-25
Recall@k Surrogate Loss with Large Batches and Similarity Mixup Code
#4ROADMAP (ResNet-50)
91.3
R@16· 2021-10-01
Robust and Decomposable Average Precision for Image Retrieval Code
#5Smooth-APSOTA
90.3
R@16· 2020-07-23
Smooth-AP: Smoothing the Path Towards Large-Scale Image Retrieval Code
#6PNP Loss
89.7
R@16· 2021-02-09
Rethinking the Optimization of Average Precision: Only Penalizing Negative Instances before Positive Ones is Enough Code