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

unsupervised single-dataset paper / code ↗

0.2566

0.1899

0.3233

0.1402

0.1485

0.3028

0.3155

0.3762

2026-04-10

Team KTH-RPL · submitted 2026-04-10 · training data: single-dataset · track: unsupervised

Notes

trained on SynFlow-4k Synthetic dataset only

Distance range 0-35 m

Metric	Aeva	AV2	nuScenes	TruckScenes	Waymo	Overall
mean Dynamic	0.2720	0.1279	0.1784	0.2523	0.1190	0.1899
mean Static	0.0189	0.0111	0.0066	0.0141	0.0076	0.0117
Threeway EPE	0.0859	0.0209	0.0148	0.2262	0.0291	0.0754
Threeway FG Dynamic	0.1959	0.0423	0.0308	0.6545	0.0765	0.2000
Threeway FG Static	0.0517	0.0129	0.0106	0.0121	0.0055	0.0186
Threeway BG Static	0.0100	0.0073	0.0029	0.0119	0.0053	0.0075
CAR Dynamic	0.1732	0.1049	0.1347	0.1131	0.1063	0.1264
CAR Static	0.0188	0.0107	0.0055	0.0161	0.0042	0.0111
OTHER_VEHICLES Dynamic	0.3969	0.1711	0.2168	0.1850	—	0.2424
OTHER_VEHICLES Static	0.0243	0.0153	0.0064	0.0110	—	0.0142
PEDESTRIAN Dynamic	0.2314	0.1441	0.2207	0.4186	0.1598	0.2349
PEDESTRIAN Static	0.0263	0.0121	0.0131	0.0158	0.0097	0.0154
WHEELED_VRU Dynamic	0.2863	0.0916	0.1416	0.2924	0.0910	0.1806
WHEELED_VRU Static	0.0160	0.0100	0.0054	0.0154	0.0126	0.0119
BACKGROUND Static	0.0094	0.0072	0.0027	0.0125	0.0039	0.0071

Distance range 35-70 m

Metric	Aeva	AV2	nuScenes	TruckScenes	Waymo	Overall
mean Dynamic	0.4804	0.1525	0.4525	0.3533	0.1780	0.3233
mean Static	0.0291	0.0158	0.0180	0.0166	0.0102	0.0179
Threeway EPE	0.1196	0.0301	0.0126	0.1563	0.0289	0.0695
Threeway FG Dynamic	0.3096	0.0613	0.0195	0.4500	0.0701	0.1821
Threeway FG Static	0.0290	0.0185	0.0101	0.0111	0.0084	0.0154
Threeway BG Static	0.0204	0.0106	0.0081	0.0079	0.0082	0.0110
CAR Dynamic	0.3225	0.1319	0.2476	0.1143	0.1118	0.1856
CAR Static	0.0281	0.0169	0.0213	0.0112	0.0078	0.0171
OTHER_VEHICLES Dynamic	0.4687	0.1541	0.3991	0.2647	—	0.3217
OTHER_VEHICLES Static	0.0300	0.0148	0.0169	0.0203	—	0.0205
PEDESTRIAN Dynamic	0.5968	0.2131	0.7730	0.6855	0.2816	0.5100
PEDESTRIAN Static	0.0385	0.0158	0.0138	0.0216	0.0111	0.0202
WHEELED_VRU Dynamic	0.5336	0.1110	0.3901	0.3488	0.1405	0.3048
WHEELED_VRU Static	0.0284	0.0195	0.0305	0.0222	0.0139	0.0229
BACKGROUND Static	0.0204	0.0119	0.0074	0.0076	0.0078	0.0110

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

Not evaluated at 35–70 m · 5 entries. The team chose to evaluate only 0–35 m (e.g. the model crops the point cloud at 51.2 m), so they are not compared here. See the 0–35 m ranking.

–

DeltaFlow_RVloss_AV2

unsupervised single-dataset paper / code ↗

—

0.2720

not evaluated

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

2026-05-01

Team IV · submitted 2026-05-01 · training data: single-dataset · track: unsupervised

Notes

Backbone: DeltaFlow Loss: RunoffVoteLoss (self-supervised) Training: Argoverse 2 only; no additional data; Paper Link: https://arxiv.org/abs/2608.18864

Distance range 0-35 m

Metric	Aeva	AV2	nuScenes	TruckScenes	Waymo	Overall
mean Dynamic	0.3273	0.1632	0.3702	0.3183	0.1813	0.2720
mean Static	0.0233	0.0164	0.0124	0.0371	0.0132	0.0205
Threeway EPE	0.1755	0.0295	0.0351	0.3420	0.0420	0.1248
Threeway FG Dynamic	0.4573	0.0492	0.0708	0.8978	0.0981	0.3146
Threeway FG Static	0.0330	0.0212	0.0175	0.0114	0.0115	0.0189
Threeway BG Static	0.0361	0.0182	0.0171	0.1168	0.0164	0.0409
CAR Dynamic	0.1946	0.1161	0.2377	0.1179	0.1111	0.1555
CAR Static	0.0182	0.0184	0.0103	0.0188	0.0086	0.0149
OTHER_VEHICLES Dynamic	0.3952	0.1541	0.3695	0.3322	—	0.3128
OTHER_VEHICLES Static	0.0188	0.0172	0.0112	0.0135	—	0.0152
PEDESTRIAN Dynamic	0.3380	0.2421	0.5619	0.4951	0.2780	0.3830
PEDESTRIAN Static	0.0233	0.0130	0.0135	0.0158	0.0110	0.0153
WHEELED_VRU Dynamic	0.3812	0.1406	0.3115	0.3280	0.1548	0.2632
WHEELED_VRU Static	0.0223	0.0161	0.0104	0.0209	0.0186	0.0177
BACKGROUND Static	0.0337	0.0170	0.0169	0.1165	0.0144	0.0397

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.

–

TeFlow-av2

unsupervised single-dataset paper / code ↗

—

0.3116

not evaluated

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

2026-04-10

Team Host · submitted 2026-04-10 · training data: single-dataset · track: unsupervised

Notes

Feed-forward SSL on av2 only, teflow-deltaflow-894989-e14-test-v3; NOTE: only 0-35m scores are valid since we set the range of pc.

Distance range 0-35 m

Metric	Aeva	AV2	nuScenes	TruckScenes	Waymo	Overall
mean Dynamic	0.3875	0.2019	0.3822	0.3884	0.1979	0.3116
mean Static	0.0288	0.0169	0.0091	0.0249	0.0118	0.0183
Threeway EPE	0.1458	0.0354	0.0344	0.3071	0.0441	0.1134
Threeway FG Dynamic	0.3646	0.0702	0.0842	0.8557	0.1080	0.2965
Threeway FG Static	0.0434	0.0218	0.0138	0.0121	0.0112	0.0205
Threeway BG Static	0.0295	0.0142	0.0052	0.0534	0.0131	0.0231
CAR Dynamic	0.2451	0.1673	0.2889	0.1584	0.1536	0.2026
CAR Static	0.0208	0.0152	0.0074	0.0175	0.0069	0.0136
OTHER_VEHICLES Dynamic	0.5092	0.2208	0.4659	0.4067	—	0.4006
OTHER_VEHICLES Static	0.0222	0.0182	0.0078	0.0144	—	0.0156
PEDESTRIAN Dynamic	0.3386	0.2734	0.4850	0.6104	0.2589	0.3933
PEDESTRIAN Static	0.0295	0.0202	0.0179	0.0280	0.0164	0.0224
WHEELED_VRU Dynamic	0.4573	0.1461	0.2891	0.3780	0.1811	0.2903
WHEELED_VRU Static	0.0444	0.0178	0.0075	0.0153	0.0133	0.0196
BACKGROUND Static	0.0272	0.0131	0.0049	0.0495	0.0106	0.0211

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.

–

VoteFlow-av2

unsupervised single-dataset

—

0.5855

not evaluated

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

2026-04-10

Team AV2 Host Team · submitted 2026-04-10 · training data: single-dataset · track: unsupervised

Notes

Feed-forward SSL on av2 only, voteflow-voteflow_ivmind_bs4_decoder_4_2-e7-test-v3; NOTE: only 0-35m scores are valid since we set the range of pc.

Distance range 0-35 m

Metric	Aeva	AV2	nuScenes	TruckScenes	Waymo	Overall
mean Dynamic	0.7961	0.2904	0.7035	0.8310	0.3067	0.5855
mean Static	0.0186	0.0161	0.0119	0.0134	0.0101	0.0140
Threeway EPE	0.2281	0.0452	0.0573	0.4277	0.0502	0.1617
Threeway FG Dynamic	0.6338	0.1036	0.1446	1.2620	0.1304	0.4549
Threeway FG Static	0.0404	0.0209	0.0192	0.0104	0.0132	0.0208
Threeway BG Static	0.0102	0.0112	0.0082	0.0108	0.0071	0.0095
CAR Dynamic	0.6445	0.2260	0.4602	0.6252	0.2139	0.4340
CAR Static	0.0189	0.0150	0.0091	0.0155	0.0110	0.0139
OTHER_VEHICLES Dynamic	0.9473	0.3012	0.7939	0.8549	—	0.7243
OTHER_VEHICLES Static	0.0210	0.0246	0.0175	0.0101	—	0.0183
PEDESTRIAN Dynamic	0.8446	0.3827	0.7983	0.9091	0.3752	0.6620
PEDESTRIAN Static	0.0272	0.0171	0.0169	0.0157	0.0135	0.0181
WHEELED_VRU Dynamic	0.7481	0.2516	0.7615	0.9348	0.3312	0.6054
WHEELED_VRU Static	0.0173	0.0132	0.0083	0.0153	0.0110	0.0130
BACKGROUND Static	0.0084	0.0103	0.0079	0.0104	0.0049	0.0084

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.

–

SeFlow++-av2

unsupervised single-dataset

—

0.5919

not evaluated

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

2026-04-10

Team Host · submitted 2026-04-10 · training data: single-dataset · track: unsupervised

Notes

Feed-forward SSL on av2 only, seflowpp-deflowpp-13987267-e5-test-v3; NOTE: only 0-35m scores are valid since we set the range of pc.

Distance range 0-35 m

Metric	Aeva	AV2	nuScenes	TruckScenes	Waymo	Overall
mean Dynamic	0.8057	0.2827	0.6965	0.8555	0.3192	0.5919
mean Static	0.0181	0.0160	0.0112	0.0132	0.0098	0.0137
Threeway EPE	0.2366	0.0420	0.0550	0.4399	0.0491	0.1645
Threeway FG Dynamic	0.6636	0.0949	0.1378	1.2993	0.1271	0.4645
Threeway FG Static	0.0358	0.0202	0.0186	0.0103	0.0125	0.0195
Threeway BG Static	0.0102	0.0108	0.0086	0.0101	0.0076	0.0095
CAR Dynamic	0.6794	0.2023	0.4641	0.6966	0.2067	0.4498
CAR Static	0.0176	0.0148	0.0106	0.0150	0.0108	0.0138
OTHER_VEHICLES Dynamic	0.9532	0.3075	0.7848	0.8937	—	0.7348
OTHER_VEHICLES Static	0.0208	0.0279	0.0118	0.0101	—	0.0176
PEDESTRIAN Dynamic	0.8372	0.3867	0.8698	0.9326	0.3579	0.6768
PEDESTRIAN Static	0.0264	0.0148	0.0169	0.0157	0.0125	0.0172
WHEELED_VRU Dynamic	0.7531	0.2343	0.6672	0.8990	0.3930	0.5893
WHEELED_VRU Static	0.0173	0.0126	0.0086	0.0154	0.0106	0.0129
BACKGROUND Static	0.0085	0.0100	0.0083	0.0098	0.0053	0.0084

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.

–

SeFlow-av2

unsupervised single-dataset paper / code ↗

—

0.6289

not evaluated

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

2026-04-10

Team AV2 Host Team · submitted 2026-04-10 · training data: single-dataset · track: unsupervised

Notes

Feed-forward SSL on av2 only; NOTE: only 0-35m scores are valid since we set the range of pc.

Distance range 0-35 m

Metric	Aeva	AV2	nuScenes	TruckScenes	Waymo	Overall
mean Dynamic	0.8412	0.3113	0.7543	0.8885	0.3494	0.6289
mean Static	0.0182	0.0167	0.0167	0.0131	0.0114	0.0152
Threeway EPE	0.2573	0.0476	0.0601	0.4627	0.0572	0.1770
Threeway FG Dynamic	0.7257	0.1093	0.1474	1.3679	0.1463	0.4993
Threeway FG Static	0.0373	0.0223	0.0224	0.0105	0.0169	0.0219
Threeway BG Static	0.0090	0.0112	0.0106	0.0097	0.0082	0.0097
CAR Dynamic	0.7046	0.2365	0.5129	0.7078	0.2334	0.4790
CAR Static	0.0186	0.0166	0.0137	0.0150	0.0143	0.0156
OTHER_VEHICLES Dynamic	0.9604	0.2949	0.8430	0.8673	—	0.7414
OTHER_VEHICLES Static	0.0208	0.0236	0.0299	0.0101	—	0.0211
PEDESTRIAN Dynamic	0.8867	0.4406	0.9291	1.0312	0.4454	0.7466
PEDESTRIAN Static	0.0263	0.0175	0.0187	0.0157	0.0150	0.0186
WHEELED_VRU Dynamic	0.8132	0.2733	0.7322	0.9477	0.3693	0.6272
WHEELED_VRU Static	0.0176	0.0153	0.0113	0.0153	0.0107	0.0140
BACKGROUND Static	0.0077	0.0103	0.0102	0.0096	0.0057	0.0087

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”.