Metric Learning On Stanford Online Products 1

Metrics

R@1

Results

Performance results of various models on this benchmark

		Paper Title	Repository
Unicom+ViT-L@336px	91.2	Unicom: Universal and Compact Representation Learning for Image Retrieval
STIR	88.3	STIR: Siamese Transformer for Image Retrieval Postprocessing
Recall@k Surrogate Loss (ViT-B/16)	88.0	Recall@k Surrogate Loss with Large Batches and Similarity Mixup
ViT-Triplet	86.5	STIR: Siamese Transformer for Image Retrieval Postprocessing
ROADMAP (DeiT-S)	86.0	Robust and Decomposable Average Precision for Image Retrieval
Hyp-ViT	85.9	Hyperbolic Vision Transformers: Combining Improvements in Metric Learning
Hyp-DINO	85.1	Hyperbolic Vision Transformers: Combining Improvements in Metric Learning
Recall@k Surrogate Loss (ViT-B/32)	85.1	Recall@k Surrogate Loss with Large Batches and Similarity Mixup
ROADMAP (ResNet-50)	83.1	Robust and Decomposable Average Precision for Image Retrieval
CCL (ResNet-50)	83.10	Center Contrastive Loss for Metric Learning	-
Recall@k Surrogate Loss (ResNet-50)	82.7	Recall@k Surrogate Loss with Large Batches and Similarity Mixup
Gradient Surgery	82.3	Dissecting the impact of different loss functions with gradient surgery	-
HAPPIER_F	81.8	Hierarchical Average Precision Training for Pertinent Image Retrieval
SCT(512)	81.6	Hard negative examples are hard, but useful
ResNet50 + Language	81.3	Integrating Language Guidance into Vision-based Deep Metric Learning
ResNet-50 + Metrix	81.3	It Takes Two to Tango: Mixup for Deep Metric Learning
NED	81.2	Calibrated neighborhood aware confidence measure for deep metric learning	-
ResNet-50 + Cross-Entropy	81.1	A unifying mutual information view of metric learning: cross-entropy vs. pairwise losses
ResNet50 + S2SD	81.0	S2SD: Simultaneous Similarity-based Self-Distillation for Deep Metric Learning
HAPPIER	81.0	Hierarchical Average Precision Training for Pertinent Image Retrieval

0 of 33 row(s) selected.

Command Palette

Metric Learning On Stanford Online Products 1

Metrics

Results