Leaderboard

Scene flow on Argoverse 2, Waymo, TruckScenes, nuScenes and Aeva, ranked by mean Dynamic (lower is better). Challenge details

Method

mean Dynamic [0-70]

mean Dynamic [0-35]

mean Dynamic [35-70]

AV2 [0-70]

Waymo [0-70]

TruckScenes [0-70]

NuScenes [0-70]

Aeva [0-70]

Date

SynFlow-4k w real-data

supervised multi-dataset paper / code ↗

0.2200

0.1602

0.2799

0.1151

0.1258

0.2775

0.3125

0.2692

2026-04-16

Team KTH-RPL · submitted 2026-04-16 · training data: multi-dataset · track: supervised

Notes

train on synflow-4k data and four real-world datasets ref. UniFlow (av2, nus, waymo, truckscene train set)

Distance range 0-35 m

Metric	Aeva	AV2	nuScenes	TruckScenes	Waymo	Overall
mean Dynamic	0.2360	0.1025	0.1405	0.2224	0.0997	0.1602
mean Static	0.0182	0.0110	0.0068	0.0132	0.0079	0.0114
Threeway EPE	0.0661	0.0173	0.0137	0.0743	0.0173	0.0377
Threeway FG Dynamic	0.1127	0.0319	0.0278	0.2043	0.0407	0.0835
Threeway FG Static	0.0788	0.0125	0.0104	0.0107	0.0057	0.0236
Threeway BG Static	0.0066	0.0074	0.0028	0.0081	0.0057	0.0061
CAR Dynamic	0.1412	0.0838	0.1237	0.0706	0.0731	0.0985
CAR Static	0.0197	0.0107	0.0056	0.0160	0.0045	0.0113
OTHER_VEHICLES Dynamic	0.3180	0.1344	0.1327	0.1563	—	0.1854
OTHER_VEHICLES Static	0.0209	0.0156	0.0071	0.0106	—	0.0135
PEDESTRIAN Dynamic	0.2135	0.1179	0.1880	0.4034	0.1436	0.2133
PEDESTRIAN Static	0.0250	0.0119	0.0124	0.0151	0.0098	0.0148
WHEELED_VRU Dynamic	0.2714	0.0739	0.1177	0.2594	0.0823	0.1610
WHEELED_VRU Static	0.0156	0.0096	0.0059	0.0158	0.0130	0.0120
BACKGROUND Static	0.0100	0.0072	0.0027	0.0086	0.0043	0.0066

Distance range 35-70 m

Metric	Aeva	AV2	nuScenes	TruckScenes	Waymo	Overall
mean Dynamic	0.3024	0.1277	0.4846	0.3327	0.1519	0.2799
mean Static	0.0222	0.0149	0.0186	0.0177	0.0127	0.0172
Threeway EPE	0.0562	0.0271	0.0135	0.0718	0.0235	0.0384
Threeway FG Dynamic	0.1154	0.0519	0.0191	0.1935	0.0496	0.0859
Threeway FG Static	0.0441	0.0179	0.0125	0.0117	0.0091	0.0190
Threeway BG Static	0.0092	0.0115	0.0090	0.0102	0.0119	0.0104
CAR Dynamic	0.1903	0.1125	0.2560	0.1046	0.0822	0.1491
CAR Static	0.0231	0.0165	0.0254	0.0117	0.0085	0.0170
OTHER_VEHICLES Dynamic	0.2773	0.1249	0.3662	0.2141	—	0.2456
OTHER_VEHICLES Static	0.0208	0.0153	0.0231	0.0201	—	0.0198
PEDESTRIAN Dynamic	0.3194	0.1780	0.8708	0.6580	0.2523	0.4557
PEDESTRIAN Static	0.0314	0.0156	0.0100	0.0219	0.0113	0.0180
WHEELED_VRU Dynamic	0.4225	0.0954	0.4454	0.3542	0.1212	0.2877
WHEELED_VRU Static	0.0273	0.0143	0.0262	0.0259	0.0199	0.0227
BACKGROUND Static	0.0086	0.0126	0.0081	0.0088	0.0111	0.0098

Imported from the original leaderboard at sceneflow.argoverse.org.

UniFlow-deltaflow

supervised multi-dataset paper / code ↗

0.2421

0.1740

0.3103

0.1242

0.1397

0.2997

0.3613

0.2856

2026-04-06

Team Host · submitted 2026-04-06 · training data: multi-dataset · track: supervised

Notes

UniFlow version trained on four datasets: av2, nus, waymo, truckscene for DeltaFlow model, with long-range auto-adaptation.

Distance range 0-35 m

Metric	Aeva	AV2	nuScenes	TruckScenes	Waymo	Overall
mean Dynamic	0.2415	0.1088	0.1678	0.2408	0.1109	0.1740
mean Static	0.0198	0.0122	0.0097	0.0150	0.0107	0.0135
Threeway EPE	0.0590	0.0187	0.0174	0.0675	0.0192	0.0363
Threeway FG Dynamic	0.1167	0.0340	0.0334	0.1777	0.0402	0.0804
Threeway FG Static	0.0508	0.0133	0.0128	0.0104	0.0081	0.0191
Threeway BG Static	0.0097	0.0087	0.0059	0.0143	0.0092	0.0095
CAR Dynamic	0.1485	0.0841	0.1374	0.0723	0.0733	0.1031
CAR Static	0.0195	0.0111	0.0085	0.0165	0.0068	0.0125
OTHER_VEHICLES Dynamic	0.3171	0.1412	0.1529	0.2047	—	0.2040
OTHER_VEHICLES Static	0.0221	0.0150	0.0094	0.0121	—	0.0147
PEDESTRIAN Dynamic	0.2162	0.1321	0.2452	0.4120	0.1558	0.2323
PEDESTRIAN Static	0.0255	0.0148	0.0156	0.0164	0.0127	0.0170
WHEELED_VRU Dynamic	0.2841	0.0776	0.1358	0.2744	0.1035	0.1751
WHEELED_VRU Static	0.0186	0.0118	0.0091	0.0158	0.0153	0.0141
BACKGROUND Static	0.0131	0.0085	0.0058	0.0142	0.0079	0.0099

Distance range 35-70 m

Metric	Aeva	AV2	nuScenes	TruckScenes	Waymo	Overall
mean Dynamic	0.3297	0.1397	0.5548	0.3586	0.1686	0.3103
mean Static	0.0214	0.0162	0.0132	0.0157	0.0153	0.0164
Threeway EPE	0.0541	0.0278	0.0134	0.0797	0.0252	0.0400
Threeway FG Dynamic	0.1275	0.0540	0.0279	0.2217	0.0544	0.0971
Threeway FG Static	0.0280	0.0191	0.0069	0.0109	0.0111	0.0152
Threeway BG Static	0.0069	0.0102	0.0053	0.0065	0.0101	0.0078
CAR Dynamic	0.1871	0.1182	0.3766	0.1070	0.0883	0.1754
CAR Static	0.0219	0.0187	0.0150	0.0125	0.0106	0.0157
OTHER_VEHICLES Dynamic	0.3047	0.1312	0.5052	0.3104	—	0.3129
OTHER_VEHICLES Static	0.0201	0.0159	0.0151	0.0183	—	0.0174
PEDESTRIAN Dynamic	0.3280	0.2032	0.7620	0.6652	0.2690	0.4455
PEDESTRIAN Static	0.0306	0.0195	0.0096	0.0200	0.0154	0.0190
WHEELED_VRU Dynamic	0.4991	0.1061	0.5753	0.3518	0.1483	0.3361
WHEELED_VRU Static	0.0279	0.0154	0.0214	0.0220	0.0260	0.0225
BACKGROUND Static	0.0064	0.0113	0.0050	0.0056	0.0094	0.0075

Imported from the original leaderboard at sceneflow.argoverse.org.

UniFlow-flow4d

supervised multi-dataset paper / code ↗

0.2644

0.1955

0.3333

0.1496

0.1608

0.3138

0.3618

0.3358

2026-04-06

Team Host · submitted 2026-04-06 · training data: multi-dataset · track: supervised

Notes

UniFlow version trained on four datasets: av2, nus, waymo, truckscene for Flow4D model, with long-range auto-adaptation.

Distance range 0-35 m

Metric	Aeva	AV2	nuScenes	TruckScenes	Waymo	Overall
mean Dynamic	0.2837	0.1184	0.2085	0.2485	0.1182	0.1955
mean Static	0.0220	0.0116	0.0071	0.0172	0.0094	0.0134
Threeway EPE	0.0781	0.0195	0.0177	0.0897	0.0191	0.0448
Threeway FG Dynamic	0.1450	0.0377	0.0392	0.2332	0.0441	0.0999
Threeway FG Static	0.0629	0.0134	0.0111	0.0103	0.0064	0.0208
Threeway BG Static	0.0263	0.0074	0.0029	0.0257	0.0069	0.0138
CAR Dynamic	0.1535	0.0953	0.1728	0.0769	0.0744	0.1146
CAR Static	0.0203	0.0109	0.0063	0.0165	0.0051	0.0119
OTHER_VEHICLES Dynamic	0.3381	0.1402	0.2496	0.2220	—	0.2375
OTHER_VEHICLES Static	0.0215	0.0155	0.0071	0.0111	—	0.0138
PEDESTRIAN Dynamic	0.2647	0.1389	0.2748	0.4122	0.1552	0.2491
PEDESTRIAN Static	0.0248	0.0136	0.0125	0.0166	0.0108	0.0156
WHEELED_VRU Dynamic	0.3785	0.0993	0.1370	0.2831	0.1249	0.2046
WHEELED_VRU Static	0.0162	0.0106	0.0066	0.0172	0.0163	0.0134
BACKGROUND Static	0.0270	0.0072	0.0028	0.0244	0.0055	0.0134

Distance range 35-70 m

Metric	Aeva	AV2	nuScenes	TruckScenes	Waymo	Overall
mean Dynamic	0.3879	0.1808	0.5151	0.3792	0.2035	0.3333
mean Static	0.0200	0.0158	0.0150	0.0160	0.0131	0.0160
Threeway EPE	0.0667	0.0329	0.0160	0.1031	0.0326	0.0503
Threeway FG Dynamic	0.1700	0.0706	0.0345	0.2928	0.0815	0.1299
Threeway FG Static	0.0256	0.0192	0.0090	0.0107	0.0089	0.0147
Threeway BG Static	0.0045	0.0088	0.0047	0.0060	0.0074	0.0063
CAR Dynamic	0.2184	0.1458	0.4622	0.1256	0.1219	0.2148
CAR Static	0.0205	0.0187	0.0201	0.0123	0.0087	0.0161
OTHER_VEHICLES Dynamic	0.3919	0.1700	0.5869	0.3502	—	0.3748
OTHER_VEHICLES Static	0.0192	0.0153	0.0201	0.0176	—	0.0181
PEDESTRIAN Dynamic	0.3804	0.2612	0.5742	0.6648	0.3037	0.4369
PEDESTRIAN Static	0.0299	0.0188	0.0101	0.0217	0.0125	0.0186
WHEELED_VRU Dynamic	0.5610	0.1461	0.4372	0.3760	0.1850	0.3411
WHEELED_VRU Static	0.0260	0.0159	0.0203	0.0241	0.0242	0.0221
BACKGROUND Static	0.0041	0.0103	0.0043	0.0045	0.0069	0.0060

Imported from the original leaderboard at sceneflow.argoverse.org.

DeltaFlow-av2

supervised single-dataset paper / code ↗

0.2688

0.1886

0.3491

0.1232

0.1444

0.3504

0.4184

0.3076

2026-04-06

Team Host · submitted 2026-04-06 · training data: single-dataset · track: supervised

Notes

only trained on av2 dataset, long-range auto-adaptation

Distance range 0-35 m

Metric	Aeva	AV2	nuScenes	TruckScenes	Waymo	Overall
mean Dynamic	0.2645	0.1033	0.1895	0.2769	0.1085	0.1886
mean Static	0.0184	0.0113	0.0076	0.0226	0.0097	0.0139
Threeway EPE	0.1311	0.0176	0.0195	0.3457	0.0292	0.1086
Threeway FG Dynamic	0.3456	0.0316	0.0427	0.9700	0.0708	0.2921
Threeway FG Static	0.0349	0.0131	0.0106	0.0112	0.0078	0.0156
Threeway BG Static	0.0127	0.0081	0.0052	0.0560	0.0090	0.0182
CAR Dynamic	0.1740	0.0838	0.1781	0.1040	0.0842	0.1248
CAR Static	0.0170	0.0115	0.0067	0.0181	0.0069	0.0120
OTHER_VEHICLES Dynamic	0.3446	0.1322	0.1756	0.2789	—	0.2328
OTHER_VEHICLES Static	0.0204	0.0165	0.0076	0.0116	—	0.0140
PEDESTRIAN Dynamic	0.2369	0.1217	0.2650	0.4160	0.1511	0.2382
PEDESTRIAN Static	0.0237	0.0114	0.0121	0.0152	0.0103	0.0145
WHEELED_VRU Dynamic	0.3026	0.0754	0.1393	0.3088	0.0903	0.1833
WHEELED_VRU Static	0.0153	0.0095	0.0064	0.0155	0.0138	0.0121
BACKGROUND Static	0.0156	0.0079	0.0051	0.0527	0.0078	0.0178

Distance range 35-70 m

Metric	Aeva	AV2	nuScenes	TruckScenes	Waymo	Overall
mean Dynamic	0.3508	0.1432	0.6473	0.4239	0.1804	0.3491
mean Static	0.0231	0.0158	0.0146	0.0224	0.0127	0.0177
Threeway EPE	0.1690	0.0297	0.0169	0.3979	0.0517	0.1330
Threeway FG Dynamic	0.4670	0.0582	0.0340	1.1633	0.1348	0.3714
Threeway FG Static	0.0266	0.0195	0.0087	0.0115	0.0105	0.0154
Threeway BG Static	0.0135	0.0115	0.0079	0.0188	0.0099	0.0123
CAR Dynamic	0.2297	0.1286	0.4808	0.1768	0.1361	0.2304
CAR Static	0.0225	0.0194	0.0186	0.0139	0.0107	0.0170
OTHER_VEHICLES Dynamic	0.3654	0.1421	0.6238	0.4458	—	0.3943
OTHER_VEHICLES Static	0.0183	0.0157	0.0175	0.0203	—	0.0179
PEDESTRIAN Dynamic	0.3328	0.1855	0.9688	0.6699	0.2683	0.4850
PEDESTRIAN Static	0.0315	0.0162	0.0110	0.0282	0.0119	0.0198
WHEELED_VRU Dynamic	0.4753	0.1167	0.5159	0.4031	0.1369	0.3296
WHEELED_VRU Static	0.0299	0.0153	0.0185	0.0299	0.0189	0.0225
BACKGROUND Static	0.0134	0.0122	0.0073	0.0198	0.0092	0.0124

Imported from the original leaderboard at sceneflow.argoverse.org.

UniFlow-ssf

supervised multi-dataset paper / code ↗

0.2968

0.2177

0.3760

0.1754

0.1867

0.3186

0.3881

0.4153

2026-04-06

Team Host · submitted 2026-04-06 · training data: multi-dataset · track: supervised

Notes

UniFlow version trained on four datasets: av2, nus, waymo, truckscene for SSF model, with long-range auto-adaptation.

Distance range 0-35 m

Metric	Aeva	AV2	nuScenes	TruckScenes	Waymo	Overall
mean Dynamic	0.3686	0.1367	0.1868	0.2597	0.1366	0.2177
mean Static	0.0184	0.0129	0.0092	0.0152	0.0118	0.0135
Threeway EPE	0.0712	0.0205	0.0188	0.0644	0.0191	0.0388
Threeway FG Dynamic	0.1619	0.0389	0.0377	0.1690	0.0417	0.0898
Threeway FG Static	0.0437	0.0147	0.0136	0.0096	0.0082	0.0179
Threeway BG Static	0.0080	0.0079	0.0052	0.0147	0.0075	0.0087
CAR Dynamic	0.1960	0.0983	0.1154	0.0778	0.0687	0.1112
CAR Static	0.0189	0.0126	0.0082	0.0169	0.0073	0.0128
OTHER_VEHICLES Dynamic	0.4697	0.1395	0.1573	0.1978	—	0.2411
OTHER_VEHICLES Static	0.0201	0.0173	0.0085	0.0118	—	0.0144
PEDESTRIAN Dynamic	0.3393	0.1851	0.3135	0.4788	0.1926	0.3018
PEDESTRIAN Static	0.0272	0.0149	0.0162	0.0165	0.0139	0.0177
WHEELED_VRU Dynamic	0.4696	0.1238	0.1612	0.2844	0.1484	0.2375
WHEELED_VRU Static	0.0141	0.0120	0.0079	0.0159	0.0190	0.0138
BACKGROUND Static	0.0115	0.0077	0.0052	0.0151	0.0069	0.0093

Distance range 35-70 m

Metric	Aeva	AV2	nuScenes	TruckScenes	Waymo	Overall
mean Dynamic	0.4620	0.2140	0.5895	0.3776	0.2369	0.3760
mean Static	0.0216	0.0200	0.0190	0.0168	0.0180	0.0191
Threeway EPE	0.0685	0.0352	0.0162	0.0795	0.0330	0.0465
Threeway FG Dynamic	0.1732	0.0733	0.0299	0.2220	0.0758	0.1148
Threeway FG Static	0.0282	0.0230	0.0133	0.0109	0.0141	0.0179
Threeway BG Static	0.0041	0.0092	0.0055	0.0057	0.0090	0.0067
CAR Dynamic	0.2564	0.1671	0.3900	0.1086	0.1087	0.2062
CAR Static	0.0219	0.0222	0.0304	0.0128	0.0135	0.0201
OTHER_VEHICLES Dynamic	0.4491	0.1397	0.4913	0.3294	—	0.3524
OTHER_VEHICLES Static	0.0198	0.0175	0.0276	0.0211	—	0.0215
PEDESTRIAN Dynamic	0.4780	0.3683	0.9322	0.6832	0.3825	0.5688
PEDESTRIAN Static	0.0335	0.0250	0.0170	0.0205	0.0225	0.0237
WHEELED_VRU Dynamic	0.6644	0.1811	0.5443	0.3891	0.2196	0.3997
WHEELED_VRU Static	0.0285	0.0247	0.0147	0.0246	0.0278	0.0241
BACKGROUND Static	0.0040	0.0104	0.0052	0.0049	0.0084	0.0066

Imported from the original leaderboard at sceneflow.argoverse.org.

SSF-long-av2

supervised single-dataset paper / code ↗

0.4691

0.4133

0.5248

0.2032

0.2518

0.6205

0.6519

0.6179

2026-04-06

Team Host · submitted 2026-04-06 · training data: single-dataset · track: supervised

Notes

trained only on av2 dataset for SSF long range config.

Distance range 0-35 m

Metric	Aeva	AV2	nuScenes	TruckScenes	Waymo	Overall
mean Dynamic	0.6300	0.1597	0.4338	0.6409	0.2020	0.4133
mean Static	0.0181	0.0142	0.0181	0.0145	0.0119	0.0154
Threeway EPE	0.2542	0.0236	0.0408	0.4834	0.0427	0.1689
Threeway FG Dynamic	0.7325	0.0471	0.0886	1.4302	0.1088	0.4815
Threeway FG Static	0.0172	0.0169	0.0230	0.0096	0.0112	0.0156
Threeway BG Static	0.0130	0.0067	0.0109	0.0103	0.0082	0.0098
CAR Dynamic	0.4224	0.1146	0.2694	0.4126	0.1262	0.2690
CAR Static	0.0166	0.0137	0.0186	0.0172	0.0095	0.0151
OTHER_VEHICLES Dynamic	0.6184	0.1699	0.3850	0.6430	—	0.4541
OTHER_VEHICLES Static	0.0206	0.0202	0.0200	0.0116	—	0.0181
PEDESTRIAN Dynamic	0.7236	0.2169	0.5549	0.7798	0.2494	0.5049
PEDESTRIAN Static	0.0256	0.0181	0.0250	0.0159	0.0171	0.0203
WHEELED_VRU Dynamic	0.7556	0.1374	0.5260	0.7283	0.2304	0.4755
WHEELED_VRU Static	0.0155	0.0127	0.0156	0.0168	0.0142	0.0149
BACKGROUND Static	0.0122	0.0065	0.0113	0.0111	0.0068	0.0096

Distance range 35-70 m

Metric	Aeva	AV2	nuScenes	TruckScenes	Waymo	Overall
mean Dynamic	0.6059	0.2467	0.8700	0.6001	0.3015	0.5248
mean Static	0.0213	0.0212	0.0205	0.0174	0.0213	0.0204
Threeway EPE	0.2334	0.0398	0.0226	0.4675	0.0730	0.1672
Threeway FG Dynamic	0.6798	0.0825	0.0514	1.3849	0.1903	0.4778
Threeway FG Static	0.0144	0.0278	0.0101	0.0114	0.0197	0.0167
Threeway BG Static	0.0060	0.0090	0.0063	0.0061	0.0091	0.0073
CAR Dynamic	0.3657	0.1859	0.8174	0.2526	0.2010	0.3645
CAR Static	0.0197	0.0281	0.0205	0.0160	0.0205	0.0210
OTHER_VEHICLES Dynamic	0.5921	0.1773	0.7532	0.5956	—	0.5296
OTHER_VEHICLES Static	0.0186	0.0216	0.0217	0.0201	—	0.0205
PEDESTRIAN Dynamic	0.6245	0.4067	0.8580	0.7857	0.4482	0.6246
PEDESTRIAN Static	0.0323	0.0260	0.0267	0.0214	0.0241	0.0261
WHEELED_VRU Dynamic	0.8412	0.2170	1.0513	0.7667	0.2552	0.6263
WHEELED_VRU Static	0.0301	0.0207	0.0276	0.0238	0.0321	0.0269
BACKGROUND Static	0.0059	0.0098	0.0061	0.0057	0.0086	0.0072

Imported from the original leaderboard at sceneflow.argoverse.org.

Not evaluated at 35–70 m · 1 entry. The team chose to evaluate only 0–35 m (e.g. the model crops the point cloud at 51.2 m), so it is not compared here. See the 0–35 m ranking.

–

UncertaintyFlow

supervised paper / code ↗

—

2.4297

not evaluated

not evaluated at 35–70 m (team's choice)

2026-06-01

Team AIEA Lab · submitted 2026-06-01 · training data: not specified · track: supervised

Distance range 0-35 m

Metric	Aeva	AV2	nuScenes	TruckScenes	Waymo	Overall
mean Dynamic	2.0177	1.8773	3.3360	2.7784	2.1388	2.4297
mean Static	0.5341	0.5429	0.5364	0.5405	0.5511	0.5410
Threeway EPE	0.7453	0.5621	0.5108	0.8713	0.6535	0.6686
Threeway FG Dynamic	1.5261	0.6377	0.5374	1.9232	0.8795	1.1008
Threeway FG Static	0.2334	0.5471	0.5157	0.1806	0.5310	0.4016
Threeway BG Static	0.4764	0.5014	0.4794	0.5102	0.5500	0.5035
CAR Dynamic	1.7925	1.3247	2.2433	1.7593	1.5783	1.7396
CAR Static	0.5487	0.5518	0.5492	0.5522	0.5509	0.5505
OTHER_VEHICLES Dynamic	1.3181	1.2463	2.5767	1.5980	—	1.6848
OTHER_VEHICLES Static	0.5417	0.5537	0.5483	0.5345	—	0.5445
PEDESTRIAN Dynamic	2.5313	3.0398	5.8072	5.3002	2.8347	3.9027
PEDESTRIAN Static	0.5561	0.5506	0.5505	0.5513	0.5511	0.5519
WHEELED_VRU Dynamic	2.4289	1.8985	2.7167	2.4562	2.0034	2.3007
WHEELED_VRU Static	0.5473	0.5511	0.5500	0.5524	0.5524	0.5506
BACKGROUND Static	0.4766	0.5075	0.4839	0.5121	0.5500	0.5060

Not evaluated at 35–70 m. The team chose to evaluate 0–35 m only (e.g. the model crops the point cloud at 51.2 m), so this entry is ranked at 0–35 m only.

Imported from the original leaderboard at sceneflow.argoverse.org.

Click a label such as supervised, unsupervised or multi-dataset to filter by it. Bold green marks the best value in each column. Dynamic metrics are speed-normalized EPE (unitless); Static and Threeway metrics are EPE in meters. Every [0-70] value is the mean of the 0–35 m and 35–70 m values (per dataset in the dataset columns); full tables are under “+ details”.