Motion Synthesis On Kit Motion Language

评估指标

Diversity

FID

Multimodality

R Precision Top3

评测结果

各个模型在此基准测试上的表现结果

模型名称	Diversity	FID	Multimodality	R Precision Top3	Paper Title	Repository
ParCo	10.95	0.453	1.245	0.772	ParCo: Part-Coordinating Text-to-Motion Synthesis	-
BAD (OAAS)	11.000	0.221	1.170	0.750	BAD: Bidirectional Auto-regressive Diffusion for Text-to-Motion Generation	-
MMM (gt length)	10.910	0.316	1.232	0.744	MMM: Generative Masked Motion Model	-
TEMOS	10.84	3.717	0.532	0.687	Executing your Commands via Motion Diffusion in Latent Space	-
T2M	10.72	3.022	2.052	0.681	Generating Diverse and Natural 3D Human Motions From Text
DisCoRD (+MoMask)	-	0.169	1.266	0.775	DisCoRD: Discrete Tokens to Continuous Motion via Rectified Flow Decoding
MotionPCM	10.827	0.294	1.254	0.787	MotionPCM: Real-Time Motion Synthesis with Phased Consistency Model	-
Motion Anything	10.94	0.131	1.374	0.802	Motion Anything: Any to Motion Generation
ReMoDiffuse	10.80	0.155	1.239	0.765	ReMoDiffuse: Retrieval-Augmented Motion Diffusion Model	-
Fg-T2M	10.93	0.571	1.019	0.745	Fg-T2M: Fine-Grained Text-Driven Human Motion Generation via Diffusion Model	-
Text2Gesture	9.334	12.12	-	0.338	TM2T: Stochastic and Tokenized Modeling for the Reciprocal Generation of 3D Human Motions and Texts	-
MMM (predict length)	10.633	0.429	1.105	0.718	MMM: Generative Masked Motion Model	-
FineMoGen	10.85	0.178	1.877	0.772	FineMoGen: Fine-Grained Spatio-Temporal Motion Generation and Editing	-
MoMask	-	0.204	1.131	0.781	MoMask: Generative Masked Modeling of 3D Human Motions	-
Free-T2M (StableMoFusion)	10.902	0.155	-	0.789	Free-T2M: Frequency Enhanced Text-to-Motion Diffusion Model With Consistency Loss	-
BiPO	10.833	0.164	1.098	0.803	BiPO: Bidirectional Partial Occlusion Network for Text-to-Motion Synthesis	-
BAMM	11.008	0.183	1.609	0.788	BAMM: Bidirectional Autoregressive Motion Model	-
DiverseMotion	10.873	0.468	2.062	0.760	DiverseMotion: Towards Diverse Human Motion Generation via Discrete Diffusion	-
MotionDiffuse	11.10	1.954	0.730	0.739	MotionDiffuse: Text-Driven Human Motion Generation with Diffusion Model	-
EMDM	10.96	0.261	1.343	0.780	EMDM: Efficient Motion Diffusion Model for Fast and High-Quality Motion Generation	-

0 of 30 row(s) selected.