Semantic Correspondence On Spair 71K

Metrics

PCK

Results

Performance results of various models on this benchmark

Model Name	PCK	Paper Title	Repository
HPF	28.2	Hyperpixel Flow: Semantic Correspondence with Multi-layer Neural Features	-
GeoAware-SC (Zero-Shot)	68.5	Telling Left from Right: Identifying Geometry-Aware Semantic Correspondence	-
GeoAware-SC (Supervised, AP-10K P.T.)	85.6	Telling Left from Right: Identifying Geometry-Aware Semantic Correspondence	-
CATs++	59.8	CATs++: Boosting Cost Aggregation with Convolutions and Transformers	-
GeoAware-SC (Supervised)	82.9	Telling Left from Right: Identifying Geometry-Aware Semantic Correspondence	-
SCOT	35.6	Semantic Correspondence as an Optimal Transport Problem
GeoAware-SC + CleanDIFT (Zero-Shot)	70.0	CleanDIFT: Diffusion Features without Noise	-
VAT	54.2	Cost Aggregation Is All You Need for Few-Shot Segmentation	-
SD+DINO (Supervised)	74.6	A Tale of Two Features: Stable Diffusion Complements DINO for Zero-Shot Semantic Correspondence	-
VAT (ECCV)	55.5	Cost Aggregation with 4D Convolutional Swin Transformer for Few-Shot Segmentation	-
SD+DINO (Zero-shot)	64.0	A Tale of Two Features: Stable Diffusion Complements DINO for Zero-Shot Semantic Correspondence	-
MMNet	50.4	Multi-scale Matching Networks for Semantic Correspondence	-
IFCAT	64.4	Integrative Feature and Cost Aggregation with Transformers for Dense Correspondence	-
DHPF	37.3	Learning to Compose Hypercolumns for Visual Correspondence	-
ANCNet	30.1	Correspondence Networks with Adaptive Neighbourhood Consensus	-
LDMCorrespondences	45.4	Unsupervised Semantic Correspondence Using Stable Diffusion	-
SD+DINO + CleanDIFT (Zero-Shot)	64.8	CleanDIFT: Diffusion Features without Noise	-
CATs	49.9	CATs: Cost Aggregation Transformers for Visual Correspondence	-
DIFT + CleanDIFT (Zero-Shot)	61.4	CleanDIFT: Diffusion Features without Noise	-
CHM	46.3	Convolutional Hough Matching Networks	-

0 of 21 row(s) selected.