T2M-GPT: Generating Human Motion from Textual Descriptions with Discrete Representations

Benchmark Model Rank Results
motion-synthesis-on-humanml3dT2M-GPT (τ = 0.5)#15FID: 0.116R Precision Top3: 0.775Diversity: 9.761
motion-synthesis-on-humanml3dT2M-GPT (τ = 0)#17FID: 0.140R Precision Top3: 0.685Diversity: 9.844
motion-synthesis-on-humanml3dT2M-GPT (τ ∈ U[0, 1])#18FID: 0.141R Precision Top3: 0.775Diversity: 9.722
motion-synthesis-on-kit-motion-languageT2M-GPT (τ ∈ U[0, 1])#17FID: 0.514R Precision Top3: 0.745Diversity: 10.921
motion-synthesis-on-kit-motion-languageT2M-GPT (τ = 0.5)#18FID: 0.717R Precision Top3: 0.737Diversity: 10.862
motion-synthesis-on-kit-motion-languageT2M-GPT (τ = 0)#19FID: 0.737R Precision Top3: 0.716Diversity: 11.198
motion-synthesis-on-motion-xT2M-GPT#2FID: 1.366TMR-R-Precision Top3: 0.655TMR-Matching Score: 0.881