Tag tests with real subjects:

Using High-Resolution Scene Camera		Using Low-Resolution Scene Camera
Scene	Results	Scene	Results
	Number of frames = 2501 Number of fixations = 160 Tag accuracy (% correct) = 98.8 Tagged video [mov] Maya visualization [mov]		Number of frames = 901 Number of fixations = 41 Tag accuracy (% correct) = 95.1 Tagged video [mov] Maya visualization [mov]
	Number of frames = 1951 Number of fixations = 148 Tag accuracy (% correct) = 98.7 Tagged video [mov] Maya visualization [mov]		Number of frames = 2201 Number of fixations = 98 Tag accuracy (% correct) = 86.7 Tagged video [mov] Maya visualization [mov]
	Number of frames = 431 Number of fixations = 48 Tag accuracy (% correct) = 97.9 Tagged video [mov] Maya visualization [mov]		Number of frames = 2901 Number of fixations = 282 Tag accuracy (% correct) = 92.0 Tagged video [mov] Maya visualization [mov]
	Number of frames = 601 Number of fixations = 59 Tag accuracy (% correct) = 98.3 Tagged video [mov] Maya visualization [mov]		Number of frames = 461 Number of fixations = 36 Tag accuracy (% correct) = 100 Tagged video [mov] Maya visualization [mov]

Tests on manually selected PORs:

The low-resolution (LR) camera outperformed the high-resolution (HR) camera because the camera intrinsic parameters of the LR camera were more accurate than those for the HR camera.

Using High-Resolution Scene Camera		Using Low-Resolution Scene Camera
Scene	Results	Scene	Results
	Number of frames = 696 Number of fixations = 121 Tag accuracy (% correct) = 87.6 POR in face mean error = 1.08in 3D PORs in Maya [png] Tagged video (1 frame/fix): Fixation Set 1 [mov] Fixation Set 2 [mov] Fixation Set 3 [mov]		Number of frames = 696 Number of fixations = 121 Tag accuracy (% correct) = 99.2 POR in face mean error = 0.54in 3D PORs in Maya [png] Tagged video (1 frame/fix): Fixation Set 1 [mov] Fixation Set 2 [mov] Fixation Set 3 [mov]

Updated May 26, 2009