Image Clustering On Cifar 10

평가 지표

ARI

Accuracy

Backbone

NMI

Train set

평가 결과

이 벤치마크에서 각 모델의 성능 결과

모델 이름	ARI	Accuracy	Backbone	NMI	Train set	Paper Title	Repository
IDFD	0.663	0.815	ResNet-18	0.711	Train+Test	Clustering-friendly Representation Learning via Instance Discrimination and Feature Decorrelation	-
VAE	0.168	0.291	VAE	0.245	Train+Test	Auto-Encoding Variational Bayes	-
IMC-SwAV (Best)	0.8	0.897	ResNet-18	0.818	Train	Information Maximization Clustering via Multi-View Self-Labelling	-
PRO-DSC	-	0.972	-	0.928	-	Exploring a Principled Framework For Deep Subspace Clustering
DEC	0.161	0.301	Custom	0.25	Train+Test	Unsupervised Deep Embedding for Clustering Analysis	-
SCAN (Avg)	0.758	0.876	ResNet-18	0.787	Train	SCAN: Learning to Classify Images without Labels	-
JULE	0.138	0.272	-	0.192	Train+Test	Joint Unsupervised Learning of Deep Representations and Image Clusters	-
GAN	0.176	0.315	GAN	0.265	Train+Test	Unsupervised Representation Learning with Deep Convolutional Generative Adversarial Networks	-
SCAN	0.772	0.883	ResNet-18	0.797	Train	SCAN: Learning to Classify Images without Labels	-
TURTLE (CLIP + DINOv2)	0.989	0.995	-	0.985	-	Let Go of Your Labels with Unsupervised Transfer	-
CoKe	0.732	0.857	ResNet-18	0.766	Train	Unsupervised Visual Representation Learning by Online Constrained K-Means	-
MMDC	-	0.820	ResNet18	0.703	-	Multi-Modal Deep Clustering: Unsupervised Partitioning of Images	-
TAC	0.831	0.919	-	0.833	-	Image Clustering with External Guidance	-
DHOG	0.492	0.666	ResNet-18	0.585	Train+Test	DHOG: Deep Hierarchical Object Grouping	-
CoHiClust	0.731	0.839	ResNet-50	0.779	Train	Contrastive Hierarchical Clustering	-
DEC+BRB	0.812	0.906	ResNet-18	0.826	Train	Breaking the Reclustering Barrier in Centroid-based Deep Clustering	-
IDEC+BRB	0.818	0.907	ResNet-18	0.833	Train	Breaking the Reclustering Barrier in Centroid-based Deep Clustering	-
RUC	-	0.903	ResNet-18	-	-	Improving Unsupervised Image Clustering With Robust Learning	-
IMC-SwAV (Avg+-)	0.79	0.891	ResNet-18	0.811	Train	Information Maximization Clustering via Multi-View Self-Labelling	-
DAC	0.301	0.522	ConvNet	0.4	Train+Test	Deep Adaptive Image Clustering

0 of 40 row(s) selected.