Table 3.

Summary of experimental set-ups and use of lameness references in studies using computer vision.

	Experimental set-up						Lameness reference
Source	Sensor type	Number of cows	Normal routine	N° of variables per variable type	Automated measurements	In real- time	Person performing scoring	Number classes used	Cut-off levels for lameness
Song et al. [34]	digital camera	15	no	1 (trackway overlap)	no	no	observer	1 to 5	all classes separate
Pluk et al. [35]	digital camera	15/66	no/yes	1 (trackway overlap)	no/yes	no/no	observer	1 to 5	all classes separate
Pluk et al. [36]	digital camera combined with Gaitwise-system	70/75	yes/yes	3 (touch angle, release angle, range of motion in the fetlock joint)	no/yes	no/no	observer	1 to 3	all classes separate
Poursaberi et al. [37]	digital camera	28/66	yes/yes	1 (back posture)	no/no	no/no	observer	1 to 3	all classes separate
Blackie et al. [64]	leg markers, digital camera	56	no	7 (stride length front and hind, tracking distance, hock flexion, max fetlock height, height of spine, head position)	no	no	observer	1 to 5	all classes separate (no 4 and 5 used)
Viazzi et al. [39]	digital camera	98	yes	1 (body movement pattern)	no	no	observer	1 to 5	non-lame: 1 + 2 lame: 3 severely lame: 4 + 5
Van Hertem et al. [43]	3D-digital camera	186	yes	1 (back posture measurement)	yes	no	observer	1 to 5	- all classes separate; - all classes separate with 1 level tolerance; - non-lame: 1 + 2 lame: 3 – 5
Viazzi et al. [40]	digital camera and 3D-digital camera	273	yes	1 back posture in 2D 1 back posture in 3D	no/yes	no/no	observer	1 to 5	non-lame: 1 + 2 lame: 3 - 5

Normal Routine: cows walk in normal routine or were guided in an experimental set up: Automated: measurements were automated without presence of an operator; real-time: results of variables are available in real-time.