Table 2.

Number of labeled samples in training, validation, and test sets for each MIM-based dataset.

		Experiment
Dataset		110		150		170		270		280		All
		P	NP	P	NP	P	NP	P	NP	P	NP	P	NP	Total
	Training	66	16	76	14	28	5	61	29	86	11	317	75	392
RAFT-MIM${}^{Q_{12}}$	Validation	13	3	15	3	5	1	13	6	18	2	64	15	79
	Test	13	3	15	3	5	1	13	6	18	2	64	15	79
	Total	92	22	106	20	38	7	87	41	122	15	445	105	550
	Training	30	6	35	6	13	1	29	13	40	4	147	30	177
RAFT-MIM${}^{Q_{25}}$	Validation	6	2	7	1	3	1	6	2	8	1	30	7	37
	Test	6	2	7	1	3	1	6	2	8	1	30	7	37
	Total	42	10	49	8	19	3	41	17	56	6	207	44	251
Farnebäck-MIM${}^{Q_{12}}$	It has the same samples as the RAFT${}^{Q_{12}}$ sets while they are generated using Farnebäck.
Farnebäck-MIM${}^{Q_{25}}$	It has the same samples as the RAFT${}^{Q_{25}}$ sets while they are generated using Farnebäck.

P: pushing samples. NP: non-pushing samples. All: all experiments. 110, 150, 170, 270, and 280: names of the video experiments.