ML Research Wiki / Benchmarks / Semantic correspondence / SPair-71k

SPair-71k

Semantic correspondence Benchmark

Performance Over Time

📊 Showing 21 results | 📏 Metric: PCK

Top Performing Models

Rank	Model	Paper	PCK	Date	Code
1	GeoAware-SC (Supervised, AP-10K P.T.)	Telling Left from Right: Identifying Geometry-Aware Semantic Correspondence	85.60	2023-11-28	📦 Junyi42/geoaware-sc
2	DINOv2	Semantic Correspondence: Unified Benchmarking and a Strong Baseline	85.20	2025-05-23	📦 Visual-AI/Semantic-Correspondence
3	GeoAware-SC (Supervised)	Telling Left from Right: Identifying Geometry-Aware Semantic Correspondence	82.90	2023-11-28	📦 Junyi42/geoaware-sc
4	SD+DINO (Supervised)	A Tale of Two Features: Stable Diffusion Complements DINO for Zero-Shot Semantic Correspondence	74.60	2023-05-24	📦 Junyi42/sd-dino
5	GeoAware-SC + CleanDIFT (Zero-Shot)	CleanDIFT: Diffusion Features without Noise	70.00	2024-12-04	📦 CompVis/cleandift
6	GeoAware-SC (Zero-Shot)	Telling Left from Right: Identifying Geometry-Aware Semantic Correspondence	68.50	2023-11-28	📦 Junyi42/geoaware-sc
7	SD+DINO + CleanDIFT (Zero-Shot)	CleanDIFT: Diffusion Features without Noise	64.80	2024-12-04	📦 CompVis/cleandift
8	IFCAT	Integrative Feature and Cost Aggregation with Transformers for Dense Correspondence	64.40	2022-09-19	-
9	SD+DINO (Zero-shot)	A Tale of Two Features: Stable Diffusion Complements DINO for Zero-Shot Semantic Correspondence	64.00	2023-05-24	📦 Junyi42/sd-dino
10	DIFT + CleanDIFT (Zero-Shot)	CleanDIFT: Diffusion Features without Noise	61.40	2024-12-04	📦 CompVis/cleandift

All Papers (21)

Telling Left from Right: Identifying Geometry-Aware Semantic Correspondence

2023

GeoAware-SC (Supervised, AP-10K P.T.)

Junyi42/geoaware-sc

Semantic Correspondence: Unified Benchmarking and a Strong Baseline

2025

DINOv2

Visual-AI/Semantic-Correspondence

Telling Left from Right: Identifying Geometry-Aware Semantic Correspondence

2023

GeoAware-SC (Supervised)

Junyi42/geoaware-sc

A Tale of Two Features: Stable Diffusion Complements DINO for Zero-Shot Semantic Correspondence

2023

SD+DINO (Supervised)

Junyi42/sd-dino

CleanDIFT: Diffusion Features without Noise

2024

GeoAware-SC + CleanDIFT (Zero-Shot)

CompVis/cleandift

Telling Left from Right: Identifying Geometry-Aware Semantic Correspondence

2023

GeoAware-SC (Zero-Shot)

Junyi42/geoaware-sc

CleanDIFT: Diffusion Features without Noise

2024

SD+DINO + CleanDIFT (Zero-Shot)

CompVis/cleandift

Integrative Feature and Cost Aggregation with Transformers for Dense Correspondence

2022

IFCAT

A Tale of Two Features: Stable Diffusion Complements DINO for Zero-Shot Semantic Correspondence

2023

SD+DINO (Zero-shot)

Junyi42/sd-dino

CleanDIFT: Diffusion Features without Noise

2024

DIFT + CleanDIFT (Zero-Shot)

CompVis/cleandift

CATs++: Boosting Cost Aggregation with Convolutions and Transformers

2022

CATs++

KU-CVLAB/CATs-PlusPlus

Cost Aggregation with 4D Convolutional Swin Transformer for Few-Shot Segmentation

2022

VAT (ECCV)

Seokju-Cho/Volumetric-Aggregation-Transformer

Cost Aggregation Is All You Need for Few-Shot Segmentation

2021

VAT

Seokju-Cho/Volumetric-Aggregation-Transformer SunghwanHong/Cost-Aggregation-transformers

TransforMatcher: Match-to-Match Attention for Semantic Correspondence

2022

TransforMatcher

wookiekim/transformatcher

Multi-scale Matching Networks for Semantic Correspondence

2021

MMNet

wintersun661/MMNet

CATs: Cost Aggregation Transformers for Visual Correspondence

2021

CATs

SunghwanHong/Cost-Aggregation-transformers

Convolutional Hough Matching Networks

2021

CHM

juhongm999/chm

Unsupervised Semantic Correspondence Using Stable Diffusion

2023

LDMCorrespondences

ubc-vision/LDM_correspondences

Learning to Compose Hypercolumns for Visual Correspondence

2020

DHPF

juhongm999/dhpf

Correspondence Networks with Adaptive Neighbourhood Consensus

2020

ANCNet

ActiveVisionLab/ANCNet

Hyperpixel Flow: Semantic Correspondence with Multi-layer Neural Features

2019

HPF

juhongm999/hpf

SPair-71k

Performance Over Time

Edit Benchmark Results

Edit Result

Top Performing Models

All Papers (21)

Telling Left from Right: Identifying Geometry-Aware Semantic Correspondence

Semantic Correspondence: Unified Benchmarking and a Strong Baseline

Telling Left from Right: Identifying Geometry-Aware Semantic Correspondence

A Tale of Two Features: Stable Diffusion Complements DINO for Zero-Shot Semantic Correspondence

CleanDIFT: Diffusion Features without Noise

Telling Left from Right: Identifying Geometry-Aware Semantic Correspondence

CleanDIFT: Diffusion Features without Noise

Integrative Feature and Cost Aggregation with Transformers for Dense Correspondence

A Tale of Two Features: Stable Diffusion Complements DINO for Zero-Shot Semantic Correspondence

CleanDIFT: Diffusion Features without Noise

CATs++: Boosting Cost Aggregation with Convolutions and Transformers

Cost Aggregation with 4D Convolutional Swin Transformer for Few-Shot Segmentation

Cost Aggregation Is All You Need for Few-Shot Segmentation

TransforMatcher: Match-to-Match Attention for Semantic Correspondence

Multi-scale Matching Networks for Semantic Correspondence

CATs: Cost Aggregation Transformers for Visual Correspondence

Convolutional Hough Matching Networks

Unsupervised Semantic Correspondence Using Stable Diffusion

Learning to Compose Hypercolumns for Visual Correspondence

Correspondence Networks with Adaptive Neighbourhood Consensus

Hyperpixel Flow: Semantic Correspondence with Multi-layer Neural Features

Model	Paper	PCK	Date
GeoAware-SC (Supervised, AP-10K P.T.)	Telling Left from Right: Identifying Geometry-Awa…	85.60	2023-11-28
DINOv2	Semantic Correspondence: Unified Benchmarking and…	85.20	2025-05-23
GeoAware-SC (Supervised)	Telling Left from Right: Identifying Geometry-Awa…	82.90	2023-11-28
SD+DINO (Supervised)	A Tale of Two Features: Stable Diffusion Compleme…	74.60	2023-05-24
GeoAware-SC + CleanDIFT (Zero-Shot)	CleanDIFT: Diffusion Features without Noise	70.00	2024-12-04
GeoAware-SC (Zero-Shot)	Telling Left from Right: Identifying Geometry-Awa…	68.50	2023-11-28
SD+DINO + CleanDIFT (Zero-Shot)	CleanDIFT: Diffusion Features without Noise	64.80	2024-12-04
IFCAT	Integrative Feature and Cost Aggregation with Tra…	64.40	2022-09-19
SD+DINO (Zero-shot)	A Tale of Two Features: Stable Diffusion Compleme…	64.00	2023-05-24
DIFT + CleanDIFT (Zero-Shot)	CleanDIFT: Diffusion Features without Noise	61.40	2024-12-04
CATs++	CATs++: Boosting Cost Aggregation with Convolutio…	59.80	2022-02-14
VAT (ECCV)	Cost Aggregation with 4D Convolutional Swin Trans…	55.50	2022-07-22
VAT	Cost Aggregation Is All You Need for Few-Shot Seg…	54.20	2021-12-22
TransforMatcher	TransforMatcher: Match-to-Match Attention for Sem…	53.70	2022-05-23
MMNet	Multi-scale Matching Networks for Semantic Corres…	50.40	2021-07-31
CATs	CATs: Cost Aggregation Transformers for Visual Co…	49.90	2021-06-04
CHM	Convolutional Hough Matching Networks	46.30	2021-03-31
LDMCorrespondences	Unsupervised Semantic Correspondence Using Stable…	45.40	2023-05-24
DHPF	Learning to Compose Hypercolumns for Visual Corre…	37.30	2020-07-21
ANCNet	Correspondence Networks with Adaptive Neighbourho…	30.10	2020-03-26
HPF	Hyperpixel Flow: Semantic Correspondence with Mul…	28.20	2019-08-18