Trajectory Prediction on ETH UCY Average
0.17ADEUPDD
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| UPDDevaluation_protocol=best-of-202026.05 | 0.17 | 0.32 | — | — | |
| Resonanceevaluation_protocol=best-of-202026.05 | 0.17 | 0.28 | — | — | |
| Reverberationevaluation_protocol=best-of-202026.05 | 0.17 | 0.28 | — | — | |
| Encoreevaluation_protocol=best-of-202026.05 | 0.17 | 0.28 | — | — | |
| CODAK (samples)=20, Evaluation Protocol=Marginal2026.05 | 0.17 | 0.28 | — | — | |
| Trajectron++Sampling=BOsampler, Strategy=Best-of-202023.04 | 0.18 | 0.36 | — | — | |
| Trajectron++Sampling=BOsampler + QMC, Strategy=Best-of-202023.04 | 0.18 | 0.36 | — | — | |
| TrajCLIPIn/Out (s)=3.2/4.8, uses_external_data=true2026.03 | 0.18 | 0.33 | — | — | |
| Ours (wide)In/Out (s)=3.2/4.82026.03 | 0.18 | 0.32 | — | — | |
| E-V2-Netevaluation_protocol=best-of-202026.05 | 0.18 | 0.3 | — | — | |
| AgentFormerevaluation_protocol=best-of-202026.05 | 0.18 | 0.29 | — | — | |
| SocialCircleevaluation_protocol=best-of-202026.05 | 0.18 | 0.29 | — | — | |
| SocialCircle+evaluation_protocol=best-of-202026.05 | 0.18 | 0.29 | — | — | |
| Y-netevaluation_protocol=best-of-202026.05 | 0.18 | 0.27 | — | — | |
| NMRFIn/Out (s)=3.2/4.82026.03 | 0.19 | 0.32 | — | — | |
| T2PIn/Out (s)=3.2/4.8, adapted_for_task=true2026.03 | 0.19 | 0.39 | — | — | |
| Ours (deep)In/Out (s)=3.2/4.82026.03 | 0.19 | 0.32 | — | — | |
| MSRLevaluation_protocol=best-of-202026.05 | 0.19 | 0.33 | — | — | |
| NRMFK (samples)=20, Evaluation Protocol=Marginal2026.05 | 0.19 | 0.32 | — | — | |
| LG-Trajevaluation_protocol=best-of-202026.05 | 0.2 | 0.34 | — | — | |
| PPTevaluation_protocol=best-of-202026.05 | 0.2 | 0.31 | — | — | |
| MoFlowK (samples)=20, Evaluation Protocol=Marginal2026.05 | 0.2 | 0.32 | — | — | |
| NPSN-SGCNsampling_strategy=NPSN2022.03 | 0.21 | 0.36 | — | — | |
| Trajectron++Sampling=MC, Strategy=Best-of-202023.04 | 0.21 | 0.41 | — | — | |
| Trajectron++Sampling=QMC, Strategy=Best-of-202023.04 | 0.21 | 0.4 | — | — | |
| MIDIn/Out (s)=3.2/4.82026.03 | 0.21 | 0.38 | — | — | |
| TUTRIn/Out (s)=3.2/4.82026.03 | 0.21 | 0.36 | — | — | |
| NPSN-NCE-Trajectron++sampling_strategy=NPSN2022.03 | 0.22 | 0.37 | — | — | |
| SingularTraj.In/Out (s)=3.2/4.82026.03 | 0.22 | 0.34 | — | — | |
| MS-TIPevaluation_protocol=best-of-202026.05 | 0.22 | 0.36 | — | — | |
| SMEMOevaluation_protocol=best-of-202026.05 | 0.22 | 0.35 | — | — | |
| MemoNetK (samples)=20, Evaluation Protocol=Marginal2026.05 | 0.22 | 0.36 | — | — | |
| GP-GraphIn/Out (s)=3.2/4.82026.03 | 0.23 | 0.39 | — | — | |
| AgentFormerK (samples)=20, Evaluation Protocol=Marginal2026.05 | 0.23 | 0.39 | — | — | |
| Trajectron++ (updated version)Sampling=BOsampler, Strategy=Best-of-202023.04 | 0.25 | 0.45 | — | — | |
| Trajectron++ (updated version)Sampling=BOsampler + QMC, Strategy=Best-of-202023.04 | 0.25 | 0.45 | — | — | |
| Joint AgentFormerK (samples)=20, Evaluation Protocol=Marginal2026.05 | 0.25 | 0.41 | — | — | |
| NPSN-Trajectron++sampling_strategy=NPSN2022.03 | 0.26 | 0.42 | — | — | |
| Y-NetK (samples)=20, Evaluation Protocol=Marginal2026.05 | 0.26 | 0.45 | — | — | |
| View VerticallyK (samples)=20, Evaluation Protocol=Marginal2026.05 | 0.27 | 0.4 | — | — | |
| NCE-Trajectron++2022.03 | 0.28 | 0.51 | — | — | |
| NPSN-STGCNNsampling_strategy=NPSN2022.03 | 0.28 | 0.45 | — | — | |
| NPSN-PECNetsampling_strategy=NPSN2022.03 | 0.28 | 0.44 | — | — | |
| Trajectron++ (updated version)Sampling=MC, Strategy=Best-of-202023.04 | 0.28 | 0.54 | — | — | |
| Trajectron++ (updated version)Sampling=QMC, Strategy=Best-of-202023.04 | 0.28 | 0.54 | — | — | |
| Joint View VerticallyK (samples)=20, Evaluation Protocol=Marginal2026.05 | 0.29 | 0.41 | — | — | |
| PECNetSampling=BOsampler, Strategy=Best-of-202023.04 | 0.3 | 0.51 | — | — | |
| PECNetSampling=BOsampler + QMC, Strategy=Best-of-202023.04 | 0.3 | 0.5 | — | — | |
| PECNetK (samples)=20, Evaluation Protocol=Marginal2026.05 | 0.3 | 0.55 | — | — | |
| Trajectron++2022.03 | 0.31 | 0.52 | — | — | |
| PECNetSampling=QMC, Strategy=Best-of-202023.04 | 0.31 | 0.54 | — | — | |
| PECNet2022.03 | 0.32 | 0.56 | — | — | |
| PECNetSampling=MC, Strategy=Best-of-202023.04 | 0.32 | 0.56 | — | — | |
| Trajectron++K (samples)=20, Evaluation Protocol=Marginal2026.05 | 0.32 | 0.55 | — | — | |
| SGCN2022.03 | 0.35 | 0.63 | — | — | |
| Social-STGCNNSampling=BOsampler + QMC, Strategy=Best-of-202023.04 | 0.37 | 0.62 | — | — | |
| NPSN-Causal-STGATsampling_strategy=NPSN2022.03 | 0.38 | 0.72 | — | — | |
| Social-STGCNNSampling=QMC, Strategy=Best-of-202023.04 | 0.39 | 0.65 | — | — | |
| Causal-STGAT2022.03 | 0.4 | 0.77 | — | — | |
| Social-STGCNNSampling=BOsampler, Strategy=Best-of-202023.04 | 0.41 | 0.69 | — | — | |
| NPSN-STGATsampling_strategy=NPSN2022.03 | 0.42 | 0.8 | — | — | |
| STGAT2022.03 | 0.43 | 0.83 | — | — | |
| Social-STGCNN2022.03 | 0.44 | 0.75 | — | — | |
| STGATSampling=BOsampler, Strategy=Best-of-202023.04 | 0.44 | 0.85 | — | — | |
| STGATSampling=BOsampler + QMC, Strategy=Best-of-202023.04 | 0.44 | 0.84 | — | — | |
| Social-STGCNNSampling=MC, Strategy=Best-of-202023.04 | 0.45 | 0.75 | — | — | |
| STGATSampling=QMC, Strategy=Best-of-202023.04 | 0.45 | 0.89 | — | — | |
| NextNumber of output samples=202019.02 | 0.46 | 1 | — | — | |
| STGATSampling=MC, Strategy=Best-of-202023.04 | 0.46 | 0.9 | — | — | |
| CGNSContext features=Deactivated2019.05 | 0.49 | 0.97 | — | — | |
| NPSN-SGANsampling_strategy=NPSN2022.03 | 0.5 | 0.96 | — | — | |
| NextNumber of output samples=12019.02 | 0.52 | 1.14 | — | — | |
| Social-GANSampling=BOsampler, Strategy=Best-of-202023.04 | 0.52 | 1.01 | — | — | |
| Social-GANSampling=BOsampler + QMC, Strategy=Best-of-202023.04 | 0.52 | 1 | — | — | |
| Social-GANSampling=MC, Strategy=Best-of-202023.04 | 0.53 | 1.05 | — | — | |
| Social-GANSampling=QMC, Strategy=Best-of-202023.04 | 0.53 | 1.03 | — | — | |
| S-GANK (samples)=20, Evaluation Protocol=Marginal2026.05 | 0.53 | 1.07 | — | — | |
| SoPhieNumber of output samples=202019.02 | 0.54 | 1.15 | — | — | |
| SoPhie2019.05 | 0.54 | 1.15 | — | — | |
| Social GAN (V)Number of output samples=202019.02 | 0.58 | 1.18 | — | — | |
| S-GAN2019.05 | 0.58 | 1.18 | — | — | |
| Social GAN (PV)Number of output samples=202019.02 | 0.61 | 1.21 | — | — | |
| S-GAN-P2019.05 | 0.61 | 1.21 | — | — | |
| Social-GAN2022.03 | 0.61 | 1.21 | — | — | |
| EMPMPIn/Out (s)=3.2/4.8, adapted_for_task=true2026.03 | 0.63 | 0.72 | — | — | |
| LSTMNumber of output samples=12019.02 | 0.7 | 1.52 | — | — | |
| Social LSTMNumber of output samples=12019.02 | 0.72 | 1.54 | — | — | |
| P-LSTM2019.05 | 0.72 | 1.53 | — | — | |
| S-LSTM2019.05 | 0.72 | 1.54 | — | — | |
| LinearNumber of output samples=12019.02 | 0.79 | 1.59 | — | — | |
| LR2019.05 | 0.79 | 1.59 | — | — | |
| CVM2019.05 | 0.82 | 1.62 | — | — | |
| Agentformer2022.03 | — | — | 0.23 | 0.39 | |
| Agentformer2023.03 | — | — | 0.23 | 0.39 | |
| AgentFormer2024.07 | — | — | 0.23 | 0.39 | |
| AgentFormerVenue=ICCV’21, K=20, Observability=agent-centric2026.05 | — | — | 0.23 | 0.39 | |
| AgentFormerParams (M)=0.592, Inference time (ms/agent)=18.072026.05 | — | — | 0.23 | 0.39 | |
| ART (Ours)2026.04 | — | — | 0.2 | 0.32 | |
| EigenTrajectory2024.07 | — | — | 0.21 | 0.34 | |
| EigenTrajectoryVenue=ICCV’23, K=20, Observability=agent-centric2026.05 | — | — | 0.22 | 0.36 |