Unsupervised Semantic Segmentation On Coco 7

Métriques

Accuracy

mIoU

Résultats

Résultats de performance de divers modèles sur ce benchmark

Nom du modèle	Accuracy	mIoU	Paper Title	Repository
SmooSeg (DINO, ViT-S/8)	63.2	26.7	-	-
IIC	21.8	-	Invariant Information Clustering for Unsupervised Image Classification and Segmentation
EQUSS (ViT-S)	53.8	25.8	Expand-and-Quantize: Unsupervised Semantic Segmentation Using High-Dimensional Space and Product Quantization	-
STEGO (ViT-B/8)	56.9	28.2	Unsupervised Semantic Segmentation by Distilling Feature Correspondences
CAUSE (ViT-B/8)	74.9	41.9	Causal Unsupervised Semantic Segmentation
PiCIE + H	49.99	14.36	PiCIE: Unsupervised Semantic Segmentation using Invariance and Equivariance in Clustering
SGSeg	55.7	-	Unsupervised Image Semantic Segmentation through Superpixels and Graph Neural Networks	-
Ours (SlotCon)	42.36	-	Self-Supervised Visual Representation Learning with Semantic Grouping
FS4	40.38	15.69	Fully Self-Supervised Learning for Semantic Segmentation	-
PriMaPs+HP (DINO ViT-S/8)	57.8	25.1	Boosting Unsupervised Semantic Segmentation with Principal Mask Proposals
PiCIE	48.1	-	PiCIE: Unsupervised Semantic Segmentation using Invariance and Equivariance in Clustering
U2Seg	63.9	30.2	Unsupervised Universal Image Segmentation
EQUSS	53.8	25.8	Expand-and-Quantize: Unsupervised Semantic Segmentation Using High-Dimensional Space and Product Quantization	-
SAN	52.0	-	Rethinking Alignment and Uniformity in Unsupervised Semantic Segmentation	-
CrOC (ViT-S/16, COCO+)	-	21.9	CrOC: Cross-View Online Clustering for Dense Visual Representation Learning
HP (ViT-S/8)	57.2	24.6	Leveraging Hidden Positives for Unsupervised Semantic Segmentation
CAUSE (DINOv2, ViT-B/14)	78.0	45.3	Causal Unsupervised Semantic Segmentation
STEGO (ViT-S/8)	-	24.5	Unsupervised Semantic Segmentation by Distilling Feature Correspondences
PriMaPs+STEGO (DINO ViT-B/8)	57.9	29.7	Boosting Unsupervised Semantic Segmentation with Principal Mask Proposals
ViCE	64.8	21.77	ViCE: Improving Dense Representation Learning by Superpixelization and Contrasting Cluster Assignment

0 of 22 row(s) selected.