Ranking

Rank	A1	A2	A3	A4	A5	A6	A7	A8	A9	A10	A11	A12	A13	A14	A15	Avg
1	38	40	43	40	43	40	47	58	64	43	48	42	36	50	38	45
2	38	43	46	41	46	40	49	65	73	48	49	43	38	52	38	47
3	44	46	50	47	56	47	52	63	70	54	54	48	46	58	46	52
4	53	54	54	52	56	55	58	70	78	60	59	57	48	61	56	58
5	53	52	52	53	55	55	54	71	84	56	60	58	51	64	57	58
6	52	56	55	51	57	53	64	73	81	61	60	57	49	61	53	59
7	51	53	58	52	64	53	67	94	132	65	64	57	53	67	53	66
8	58	59	64	62	65	60	68	77	92	65	68	62	60	70	59	66

• A1-A15 denotes the motion scenario index described in the Human3.6M dataset.
      A1 - Directions; A2 - Discussion; A3 - Eating; A4 - Greeting; A5 - Phone Call
      A6 - Posing; A7 - Purchases; A8 - Sitting; A9 - Sitting Down; A10- Smoking
      A11- Taking Photo; A12- Waiting; A13- Walking; A14- Walking Dog; A15- Walking Together

• CHALL_H80K is the Human80K testset (25K examples) for the ECCV 2018 PoseTrack Challenge

Challenge Participants

1) Istvan Sarandi

2) Xiao Sun

3) Georgios Pavlakos

4) Shen Yuejia

5) Sungheon Park

6) Zhou Ye Jiang

7) Helge Rhodin

8) Tyler Zhu

Rank	A1	A2	A3	A4	A5	A6	A7	A8	A9	A10	A11	A12	A13	A14	A15	Avg

• A1-A15 denotes the motion scenario index described in the Human3.6M dataset.
      A1 - Directions; A2 - Discussion; A3 - Eating; A4 - Greeting; A5 - Phone Call
      A6 - Posing; A7 - Purchases; A8 - Sitting; A9 - Sitting Down; A10- Smoking
      A11- Taking Photo; A12- Waiting; A13- Walking; A14- Walking Dog; A15- Walking Together

• CHALL_H80K is the Human80K testset (25K examples) for the ECCV 2018 PoseTrack Challenge

Challenge Participants

Rank	A1	A2	A3	A4	A5	A6	A7	A8	A9	A10	A11	A12	A13	A14	A15	Avg
1	54	53	53	57	61	55	64	80	102	60	67	55	54	68	58	63
2	76	70	74	84	77	77	107	118	156	77	110	83	76	109	75	92
3	95	83	85	99	93	91	116	124	175	91	116	95	91	115	97	105
4	127	112	109	135	121	131	146	164	229	121	149	130	112	150	132	139

• A1-A15 denotes the motion scenario index described in the Human3.6M dataset.
      A1 - Directions; A2 - Discussion; A3 - Eating; A4 - Greeting; A5 - Phone Call
      A6 - Posing; A7 - Purchases; A8 - Sitting; A9 - Sitting Down; A10- Smoking
      A11- Taking Photo; A12- Waiting; A13- Walking; A14- Walking Dog; A15- Walking Together

• H80K is the Human80K testset (25K examples)

Description

1) A.-I. Popa, M. Zanfir, C. Sminchisescu. Deep Multitask Architecture for Integrated 2D and 3D Human Sensing. IEEE Computer Vision and Pattern Recognition 2017.

2) Kernel dependency estimation regressor on O2PL features i.e. KDE-O2LP-rlpRF method from the paper:
C. Ionescu, J. Carreira, and C. Sminchisescu. Iterated Second-Order Label Sensitive Pooling for 3D Human Pose Estimation. IEEE Computer Vision and Pattern Recognition 2014.
See also our project page for details.

3) Ridge regressor (RR) on O2PL features i.e. RR-O2LP-rlpRF method from the paper:
C. Ionescu, J. Carreira, and C. Sminchisescu. Iterated Second-Order Label Sensitive Pooling for 3D Human Pose Estimation. IEEE Computer Vision and Pattern Recognition 2014.

4) Ridge regressor (RR) on exponentiated chi square (approximated with random features) of HoG features i.e. RR-exp(chi2) from the paper:
C. Ionescu, J. Carreira, and C. Sminchisescu. Iterated Second-Order Label Sensitive Pooling for 3D Human Pose Estimation. IEEE Computer Vision and Pattern Recognition 2014.

Rank	A1	A2	A3	A4	A5	A6	A7	A8	A9	A10	A11	A12	A13	A14	A15	Avg

• A1-A15 denotes the motion scenario index described in the Human3.6M dataset.
      A1 - Directions; A2 - Discussion; A3 - Eating; A4 - Greeting; A5 - Phone Call
      A6 - Posing; A7 - Purchases; A8 - Sitting; A9 - Sitting Down; A10- Smoking
      A11- Taking Photo; A12- Waiting; A13- Walking; A14- Walking Dog; A15- Walking Together

• H80K is the Human80K testset (25K examples)

Description

Rank	A1	A2	A3	A4	A5	A6	A7	A8	A9	A10	A11	A12	A13	A14	A15	Avg
1	31	33	41	34	41	37	37	51	56	43	44	37	33	42	32	39
2	40	36	44	39	44	42	41	66	70	46	49	43	34	46	34	45
3	42	44	52	47	54	48	49	66	76	54	61	47	44	55	44	52
4	51	50	54	54	62	57	54	72	76	62	65	59	49	61	54	58
5	49	47	51	52	60	56	56	82	94	64	69	61	48	66	49	60
6	54	54	63	59	72	61	68	101	109	74	81	62	55	75	60	69
7	60	56	68	64	78	67	68	106	119	77	85	64	57	78	62	73
8	91	89	94	102	105	99	112	151	239	109	151	106	101	141	106	119
9	117	108	91	129	104	130	134	135	200	117	195	132	115	162	156	133
10	123	122	121	140	125	148	157	162	212	132	205	147	124	176	158	148
11	133	133	130	150	137	157	168	173	239	147	219	160	135	194	172	161
12	152	153	125	171	135	180	162	168	221	160	241	176	157	201	187	171

• A1-A15 denotes the motion scenario index described in the Human3.6M dataset.
      A1 - Directions; A2 - Discussion; A3 - Eating; A4 - Greeting; A5 - Phone Call
      A6 - Posing; A7 - Purchases; A8 - Sitting; A9 - Sitting Down; A10- Smoking
      A11- Taking Photo; A12- Waiting; A13- Walking; A14- Walking Dog; A15- Walking Together

• H36M_NOS10 is the Human3.6M testset without subject S10 (about 800K examples).

Description

1) D.C. Luvizon, H. Tabia, D. Picard Consensus-based Optimization for 3D Human Pose Estimation in Camera Coordinates, preprint arXiv:1911.09245 (2019), *multi-view GT calibration*

2) D.C. Luvizon, H. Tabia, D. Picard Consensus-based Optimization for 3D Human Pose Estimation in Camera Coordinates, preprint arXiv:1911.09245 (2019), *multi-view estimated calibration*

3) D.C. Luvizon, H. Tabia, D. Picard Consensus-based Optimization for 3D Human Pose Estimation in Camera Coordinates, preprint arXiv:1911.09245 (2019), *monocular*

4) 2) Anonymous NIPS18 submission, PaperID: 494, FBI-Pose: Towards Bridging the Gap between 2D Images and 3D Human Poses using Forward-or-Backward Information(Update)

5) Anonymous NIPS18 submission, PaperID: 5104, Deep Network for the Integrated 3D Sensing of Multiple People in Natural Images

6) A. Zanfir, E. Marinoiu, C. Sminchisescu Monocular 3D Pose and Shape Estimation of Multiple People in Natural Scenes - The Importance of Multiple Scene Constraints, IEEE Computer Vision and Pattern Recognition 2018.

7) A.-I. Popa, M. Zanfir, C. Sminchisescu. Deep Multitask Architecture for Integrated 2D and 3D Human Sensing. IEEE Computer Vision and Pattern Recognition 2017.

8) 3D Convolutional Neural Network. The details of the method can be found in the paper: Agne Grinciunaite, Amogh Gudi, Emrah Tasli, Marten den Uyl, Human Pose Estimation in Space and Time using 3D CNN, *arXiv preprint arXiv:1609.00036* (2016).

9) Linear (random feature) approximation of the kernel dependency estimation (LinKDE) method using a pyramid of SIFTs extracted on images on which a background subtraction mask (BS) was applied. The details of the method can be found in the paper:
Catalin Ionescu, Dragos Papava, Vlad Olaru and Cristian Sminchisescu, Human3.6M: Large Scale Datasets and Predictive Methods for 3D Human Sensing in Natural Environments, IEEE Transactions on Pattern Analysis and Machine Intelligence, 2014.

10) Kernel ridge regression (KRR) method using a pyramid of SIFTs extracted on images on which a background subtraction mask (BS) was applied. The details of the method can be found in the paper:
Catalin Ionescu, Dragos Papava, Vlad Olaru and Cristian Sminchisescu, Human3.6M: Large Scale Datasets and Predictive Methods for 3D Human Sensing in Natural Environments, IEEE Transactions on Pattern Analysis and Machine Intelligence, 2014

11) Linear (random feature) approximation of the kernel ridge regression (LinKRR) method using a pyramid of SIFTs extracted on images on which a background subtraction mask (BS) was applied. The details of the method can be found in the paper:
Catalin Ionescu, Dragos Papava, Vlad Olaru and Cristian Sminchisescu, Human3.6M: Large Scale Datasets and Predictive Methods for 3D Human Sensing in Natural Environments, IEEE Transactions on Pattern Analysis and Machine Intelligence, 2014.

12) Nearest neighbor (KNN) method using a pyramid of SIFTs extracted on images on which a background subtraction mask (BS) was applied. The details of the method can be found in the paper:
Catalin Ionescu, Dragos Papava, Vlad Olaru and Cristian Sminchisescu, Human3.6M: Large Scale Datasets and Predictive Methods for 3D Human Sensing in Natural Environments, IEEE Transactions on Pattern Analysis and Machine Intelligence, 2014.

Rank	A1	A2	A3	A4	A5	A6	A7	A8	A9	A10	A11	A12	A13	A14	A15	Avg

• A1-A15 denotes the motion scenario index described in the Human3.6M dataset.
      A1 - Directions; A2 - Discussion; A3 - Eating; A4 - Greeting; A5 - Phone Call
      A6 - Posing; A7 - Purchases; A8 - Sitting; A9 - Sitting Down; A10- Smoking
      A11- Taking Photo; A12- Waiting; A13- Walking; A14- Walking Dog; A15- Walking Together

• H36M_NOS10 is the Human3.6M testset without subject S10 (about 800K examples).

Description

Rank	A1	A2	A3	A4	A5	A6	A7	A8	A9	A10	A11	A12	A13	A14	A15	Avg
1	123	115	95	141	111	144	136	140	200	116	197	145	114	164	150	138
2	131	130	119	145	130	157	158	162	215	135	200	155	120	177	151	151
3	140	139	130	156	136	167	169	174	239	145	213	163	130	190	161	162
4	164	166	135	188	143	198	173	176	234	167	245	194	160	216	189	182

• A1-A15 denotes the motion scenario index described in the Human3.6M dataset.
      A1 - Directions; A2 - Discussion; A3 - Eating; A4 - Greeting; A5 - Phone Call
      A6 - Posing; A7 - Purchases; A8 - Sitting; A9 - Sitting Down; A10- Smoking
      A11- Taking Photo; A12- Waiting; A13- Walking; A14- Walking Dog; A15- Walking Together

• H36M is the full Human3.6M test set (1.2 million examples).

Description

1) Linear (random feature) approximation of the kernel dependency estimation (LinKDE) method using a pyramid of SIFTs extracted on images on which a background subtraction mask (BS) was applied. The details can be found in the paper:
Catalin Ionescu, Dragos Papava, Vlad Olaru and Cristian Sminchisescu, Human3.6M: Large Scale Datasets and Predictive Methods for 3D Human Sensing in Natural Environments, IEEE Transactions on Pattern Analysis and Machine Intelligence, 2014

2) Kernel ridge regression (KRR) method using a pyramid of SIFTs extracted on images on which a background subtraction mask (BS) was applied. The details can be found in the paper:
Catalin Ionescu, Dragos Papava, Vlad Olaru and Cristian Sminchisescu, Human3.6M: Large Scale Datasets and Predictive Methods for 3D Human Sensing in Natural Environments, IEEE Transactions on Pattern Analysis and Machine Intelligence, 2014

3) Linear (random feature) approximation of the kernel ridge regression (LinKRR) method using a pyramid of SIFTs extracted on images on which a background subtraction mask (BS) was applied. The details can be found in the paper:
Catalin Ionescu, Dragos Papava, Vlad Olaru and Cristian Sminchisescu, Human3.6M: Large Scale Datasets and Predictive Methods for 3D Human Sensing in Natural Environments, IEEE Transactions on Pattern Analysis and Machine Intelligence, 2014

4) Nearest neighbor (KNN) method using a pyramid of SIFTs extracted on images on which a background subtraction mask (BS) was applied. The details can be found in the paper:
Catalin Ionescu, Dragos Papava, Vlad Olaru and Cristian Sminchisescu, Human3.6M: Large Scale Datasets and Predictive Methods for 3D Human Sensing in Natural Environments, IEEE Transactions on Pattern Analysis and Machine Intelligence, 2014

Rank	A1	A2	A3	A4	A5	A6	A7	A8	A9	A10	A11	A12	A13	A14	A15	Avg

• A1-A15 denotes the motion scenario index described in the Human3.6M dataset.
      A1 - Directions; A2 - Discussion; A3 - Eating; A4 - Greeting; A5 - Phone Call
      A6 - Posing; A7 - Purchases; A8 - Sitting; A9 - Sitting Down; A10- Smoking
      A11- Taking Photo; A12- Waiting; A13- Walking; A14- Walking Dog; A15- Walking Together

• H36M is the full Human3.6M test set (1.2 million examples).

Description

Rank	MR1	MR2	MR3	MR4	MR5	MR6	MR7	Avg

• MR1-MR7 denote the set of 7 mixed reality videos available for download.

Description

Rank	MR1	MR2	MR3	MR4	MR5	MR6	MR7	Avg

• MR1-MR7 denote the set of 7 mixed reality videos available for download.

Description

Rank	acc_class_all	acc_class_silh	acc_pixel_all	acc_pixel_silh
1	0.74	0.73	0.92	0.81

• These are the rankings for the human part labeling task on H80K under different accuracy metrics
• H80K is the Human80K testset (25K examples)

Description

1) A.-I. Popa, M. Zanfir, C. Sminchisescu. Deep Multitask Architecture for Integrated 2D and 3D Human Sensing. IEEE Computer Vision and Pattern Recognition 2017.

Rank	acc_class_all	acc_class_silh	acc_pixel_all	acc_pixel_silh
1	0.74	0.73	0.92	0.81

• These are the rankings for the human part labeling task on H80K under different accuracy metrics
• H80K is the Human80K testset (25K examples)

Description

1) A.-I. Popa, M. Zanfir, C. Sminchisescu. Deep Multitask Architecture for Integrated 2D and 3D Human Sensing. IEEE Computer Vision and Pattern Recognition 2017.

Rank	A1	A2	A3	A4	A5	A6	A7	A8	A9	A10	A11	A12	A13	A14	A15	Avg
1	38	40	43	40	43	40	47	58	64	43	48	42	36	50	38	45
2	38	43	46	41	46	40	49	65	73	48	49	43	38	52	38	47
3	44	46	50	47	56	47	52	63	70	54	54	48	46	58	46	52
4	53	54	54	52	56	55	58	70	78	60	59	57	48	61	56	58
5	53	52	52	53	55	55	54	71	84	56	60	58	51	64	57	58
6	52	56	55	51	57	53	64	73	81	61	60	57	49	61	53	59
7	51	53	58	52	64	53	67	94	132	65	64	57	53	67	53	66
8	58	59	64	62	65	60	68	77	92	65	68	62	60	70	59	66

Rank	A1	A2	A3	A4	A5	A6	A7	A8	A9	A10	A11	A12	A13	A14	A15	Avg
1	54	53	53	57	61	55	64	80	102	60	67	55	54	68	58	63
2	76	70	74	84	77	77	107	118	156	77	110	83	76	109	75	92
3	95	83	85	99	93	91	116	124	175	91	116	95	91	115	97	105
4	127	112	109	135	121	131	146	164	229	121	149	130	112	150	132	139

Rank	A1	A2	A3	A4	A5	A6	A7	A8	A9	A10	A11	A12	A13	A14	A15	Avg
1	31	33	41	34	41	37	37	51	56	43	44	37	33	42	32	39
2	40	36	44	39	44	42	41	66	70	46	49	43	34	46	34	45
3	42	44	52	47	54	48	49	66	76	54	61	47	44	55	44	52
4	51	50	54	54	62	57	54	72	76	62	65	59	49	61	54	58
5	49	47	51	52	60	56	56	82	94	64	69	61	48	66	49	60
6	54	54	63	59	72	61	68	101	109	74	81	62	55	75	60	69
7	60	56	68	64	78	67	68	106	119	77	85	64	57	78	62	73
8	91	89	94	102	105	99	112	151	239	109	151	106	101	141	106	119
9	117	108	91	129	104	130	134	135	200	117	195	132	115	162	156	133
10	123	122	121	140	125	148	157	162	212	132	205	147	124	176	158	148
11	133	133	130	150	137	157	168	173	239	147	219	160	135	194	172	161
12	152	153	125	171	135	180	162	168	221	160	241	176	157	201	187	171

Rank	A1	A2	A3	A4	A5	A6	A7	A8	A9	A10	A11	A12	A13	A14	A15	Avg
1	123	115	95	141	111	144	136	140	200	116	197	145	114	164	150	138
2	131	130	119	145	130	157	158	162	215	135	200	155	120	177	151	151
3	140	139	130	156	136	167	169	174	239	145	213	163	130	190	161	162
4	164	166	135	188	143	198	173	176	234	167	245	194	160	216	189	182

Rank	A1	A2	A3	A4	A5	A6	A7	A8	A9	A10	A11	A12	A13	A14	A15	Avg
1	38	40	43	40	43	40	47	58	64	43	48	42	36	50	38	45
2	38	43	46	41	46	40	49	65	73	48	49	43	38	52	38	47
3	44	46	50	47	56	47	52	63	70	54	54	48	46	58	46	52
4	53	54	54	52	56	55	58	70	78	60	59	57	48	61	56	58
5	53	52	52	53	55	55	54	71	84	56	60	58	51	64	57	58
6	52	56	55	51	57	53	64	73	81	61	60	57	49	61	53	59
7	51	53	58	52	64	53	67	94	132	65	64	57	53	67	53	66
8	58	59	64	62	65	60	68	77	92	65	68	62	60	70	59	66

Rank	A1	A2	A3	A4	A5	A6	A7	A8	A9	A10	A11	A12	A13	A14	A15	Avg
1	54	53	53	57	61	55	64	80	102	60	67	55	54	68	58	63
2	76	70	74	84	77	77	107	118	156	77	110	83	76	109	75	92
3	95	83	85	99	93	91	116	124	175	91	116	95	91	115	97	105
4	127	112	109	135	121	131	146	164	229	121	149	130	112	150	132	139

Rank	A1	A2	A3	A4	A5	A6	A7	A8	A9	A10	A11	A12	A13	A14	A15	Avg
1	31	33	41	34	41	37	37	51	56	43	44	37	33	42	32	39
2	40	36	44	39	44	42	41	66	70	46	49	43	34	46	34	45
3	42	44	52	47	54	48	49	66	76	54	61	47	44	55	44	52
4	51	50	54	54	62	57	54	72	76	62	65	59	49	61	54	58
5	49	47	51	52	60	56	56	82	94	64	69	61	48	66	49	60
6	54	54	63	59	72	61	68	101	109	74	81	62	55	75	60	69
7	60	56	68	64	78	67	68	106	119	77	85	64	57	78	62	73
8	91	89	94	102	105	99	112	151	239	109	151	106	101	141	106	119
9	117	108	91	129	104	130	134	135	200	117	195	132	115	162	156	133
10	123	122	121	140	125	148	157	162	212	132	205	147	124	176	158	148
11	133	133	130	150	137	157	168	173	239	147	219	160	135	194	172	161
12	152	153	125	171	135	180	162	168	221	160	241	176	157	201	187	171

Rank	A1	A2	A3	A4	A5	A6	A7	A8	A9	A10	A11	A12	A13	A14	A15	Avg
1	123	115	95	141	111	144	136	140	200	116	197	145	114	164	150	138
2	131	130	119	145	130	157	158	162	215	135	200	155	120	177	151	151
3	140	139	130	156	136	167	169	174	239	145	213	163	130	190	161	162
4	164	166	135	188	143	198	173	176	234	167	245	194	160	216	189	182

Results

Challenge Participants

Challenge Participants

Description

Description

Description

Description

Description

Description

Description

Description

Description

Description

Rank	A1	A2	A3	A4	A5	A6	A7	A8	A9	A10	A11	A12	A13	A14	A15	Avg
1	38	40	43	40	43	40	47	58	64	43	48	42	36	50	38	45
2	38	43	46	41	46	40	49	65	73	48	49	43	38	52	38	47
3	44	46	50	47	56	47	52	63	70	54	54	48	46	58	46	52
4	53	54	54	52	56	55	58	70	78	60	59	57	48	61	56	58
5	53	52	52	53	55	55	54	71	84	56	60	58	51	64	57	58
6	52	56	55	51	57	53	64	73	81	61	60	57	49	61	53	59
7	51	53	58	52	64	53	67	94	132	65	64	57	53	67	53	66
8	58	59	64	62	65	60	68	77	92	65	68	62	60	70	59	66

Rank	A1	A2	A3	A4	A5	A6	A7	A8	A9	A10	A11	A12	A13	A14	A15	Avg
1	54	53	53	57	61	55	64	80	102	60	67	55	54	68	58	63
2	76	70	74	84	77	77	107	118	156	77	110	83	76	109	75	92
3	95	83	85	99	93	91	116	124	175	91	116	95	91	115	97	105
4	127	112	109	135	121	131	146	164	229	121	149	130	112	150	132	139

Rank	A1	A2	A3	A4	A5	A6	A7	A8	A9	A10	A11	A12	A13	A14	A15	Avg
1	31	33	41	34	41	37	37	51	56	43	44	37	33	42	32	39
2	40	36	44	39	44	42	41	66	70	46	49	43	34	46	34	45
3	42	44	52	47	54	48	49	66	76	54	61	47	44	55	44	52
4	51	50	54	54	62	57	54	72	76	62	65	59	49	61	54	58
5	49	47	51	52	60	56	56	82	94	64	69	61	48	66	49	60
6	54	54	63	59	72	61	68	101	109	74	81	62	55	75	60	69
7	60	56	68	64	78	67	68	106	119	77	85	64	57	78	62	73
8	91	89	94	102	105	99	112	151	239	109	151	106	101	141	106	119
9	117	108	91	129	104	130	134	135	200	117	195	132	115	162	156	133
10	123	122	121	140	125	148	157	162	212	132	205	147	124	176	158	148
11	133	133	130	150	137	157	168	173	239	147	219	160	135	194	172	161
12	152	153	125	171	135	180	162	168	221	160	241	176	157	201	187	171

Rank	A1	A2	A3	A4	A5	A6	A7	A8	A9	A10	A11	A12	A13	A14	A15	Avg
1	123	115	95	141	111	144	136	140	200	116	197	145	114	164	150	138
2	131	130	119	145	130	157	158	162	215	135	200	155	120	177	151	151
3	140	139	130	156	136	167	169	174	239	145	213	163	130	190	161	162
4	164	166	135	188	143	198	173	176	234	167	245	194	160	216	189	182