Table 2.

Numerical comparisons on the EE3D-R dataset (in mm). “$\sigma$” denotes the standard deviation of MPJPE or PA-MPJPE across actions. Our EventEgo3D++ outperforms existing approaches on most activities by a substantial margin and achieves $11\%$ improvement over Rudnev et al. (2021).

Method	Metric	Walk	Crouch	Pushup	Boxing	Kick	Dance	Inter. with env.	Crawl	Sports	Jump	Avg. ($\sigma$)
Tome et al. (2019)	MPJPE	140.34	173.93	157.29	177.07	181.12	212.61	169.80	144.80	207.56	165.57	173.01 (23.62)
	PA-MPJPE	104.34	119.89	102.39	124.28	121.64	132.86	111.89	88.94	120.15	110.32	113.67 (12.76)
Xu et al. (2019)	MPJPE	86.09	153.53	199.34	133.15	114.00	104.44	114.52	187.95	128.21	114.10	133.53 (36.42)
	PA-MPJPE	59.11	113.31	147.13	102.50	91.75	79.65	85.83	138.12	98.10	89.19	100.47 (26.52)
Rudnev et al. (2021)	MPJPE	74.82	178.23	105.68	128.93	112.45	98.14	110.05	120.51	110.16	106.19	114.52 (26.54)
	PA-MPJPE	56.77	108.34	84.15	100.39	91.84	78.16	74.62	83.47	84.83	86.09	84.87 (14.08)
Millerdurai et al. (2024a)	MPJPE	70.88	163.84	97.88	136.57	103.72	88.87	103.19	109.71	101.02	97.32	107.30 (25.78)
	PA-MPJPE	52.11	99.48	75.53	104.66	86.05	71.96	70.85	77.94	77.82	80.17	79.66 (14.83)
EventEgo3D++ (Ours)	MPJPE	68.67	157.41	88.63	123.57	102.31	84.95	95.73	109.38	94.9	95.94	102.15 (23.01)
	PA-MPJPE	50.06	100.76	66.29	94.52	84.26	66.91	68.2	75.73	72.23	75.83	75.48 (13.95)