HyperAI초신경

홈 뉴스 최신 연구 논문 튜토리얼 데이터셋 백과사전 SOTA LLM 모델 GPU 랭킹 컨퍼런스

한국어

HyperAI초신경

Image Clustering On Imagenet

평가 지표

ARI

Accuracy

NMI

평가 결과

이 벤치마크에서 각 모델의 성능 결과

모델 이름	ARI	Accuracy	NMI	Paper Title	Repository
MIM-Refiner (D2V2-ViT-H/14)	42.2	67.3	87.2	MIM-Refiner: A Contrastive Learning Boost from Intermediate Pre-Trained Representations
TEMI DINO (ViT-B)	45.9	58.0	81.4	Exploring the Limits of Deep Image Clustering using Pretrained Models
TURTLE (CLIP + DINOv2)	62.5	72.9	88.2	Let Go of Your Labels with Unsupervised Transfer
MIM-Refiner (MAE-ViT-H/14)	45.5	64.6	85.3	MIM-Refiner: A Contrastive Learning Boost from Intermediate Pre-Trained Representations
TEMI MSN (ViT-L)	48.4	61.6	82.5	Exploring the Limits of Deep Image Clustering using Pretrained Models
MAE-CT (ViT-H/16 best)	-	58.0	81.8	Contrastive Tuning: A Little Help to Make Masked Autoencoders Forget
SCAN	-	39.9	72.0	SCAN: Learning to Classify Images without Labels
SeLa	-	-	66.4	Self-labelling via simultaneous clustering and representation learning
SeCu	41.9	53.5	79.4	Stable Cluster Discrimination for Deep Clustering
CoKe	35.6	47.6	76.2	Stable Cluster Discrimination for Deep Clustering
PRO-DSC	-	65.0	83.4	Exploring a Principled Framework For Deep Subspace Clustering
MAE-CT (ViT-H/16 mean)	-	57.1	81.7	Contrastive Tuning: A Little Help to Make Masked Autoencoders Forget

0 of 12 row(s) selected.