Understanding face detection with visual arrays and real-world scenes

Figures & data

Figure 1 . Example stimuli for the blank (top), array (middle) and scene (bottom) conditions, with insets in the blank display illustrating frontal (F) and profile (P) face targets.

This Figure contains three types of display presented in vertical arrangement. The display at the top is a plain white background with a frontal face in the bottom left quadrant. The display also has an inset in the top left quadrant which separately illustrates frontal and profile face targets. The display in the middle shows an array of 24 items, such as an apple, a camera or a lightbulb, which are spaced equally on a white background. The display at the bottom is a photograph of a living room setting. Four armchairs are visible, along with a fireplace and a mantelpiece with vases and other display items. Bookshelves to the left and the right of the room are also visible as well as some paintings on the wall.

Table 1 . Mean RTs (ms), Accuracy (%) and Inverse Efficiency Scores (IES) for Frontal (F) and Profile (P) Faces in the Blank, Array and Scene Conditions in Experiment 1. Parentheses Show the Standard Deviation of the Mean.

	RTs			Accuracy			IES
	Frontal	Profile	F – P	Frontal	Profile	F – P	Frontal	Profile	F – P
Blank	559 (93)	545 (80)	14	96.0 (5.42)	96.4 (4.13)	−0.4	584 (100)	566 (82)	18
Array	732 (117)	762 (136)	−30	94.9 (7.15)	93.2 (6.38)	1.7	773 (118)	818 (137)	−45
Scene	768 (132)	908 (146)	−140	91.6 (8.31)	83.9 (11.10)	7.7	847 (167)	1103 (237)	−256

Table 2 . Mean RTs (ms), Accuracy (%) and IES for Upright (U) and Rotated (R) Faces in the Blank, Array and Scene Conditions in Experiment 2. Parentheses Show the Standard Deviation of the Mean.

	RTs			Accuracy			IES
	Upright	Rotated	U – R	Upright	Rotated	U – R	Upright	Rotated	U – R
Blank	556 (105)	562 (102)	−6	95.8 (7.37)	94.3 (6.92)	1.5	583 (111)	596 (105)	−13
Array	775 (158)	889 (180)	−114	92.9 (9.75)	84.7 (13.41)	8.2	842 (183)	1083 (308)	−241
Scene	811 (202)	954 (196)	−143	89.2 (12.0)	83.1 (17.64)	6.1	922 (246)	1214 (446)	−292

Table 3 . Mean RTs (ms), Accuracy (%) and IES for Upright (U) and Inverted (I) Faces in the Blank, Array and Scene Conditions in Experiment 3. Parentheses Show the Standard Deviation of the Mean.

	RTs			Accuracy			IES
	Upright	Inverted	U – I	Upright	Inverted	U – I	Upright	Inverted	U – I
Blank	557 (89)	568 (89)	−11	96.9 (3.38)	96.4 (3.77)	0.5	575 (89)	589 (84)	−14
Array	777 (125)	889 (153)	−112	95.6 (8.54)	91.0 (10.4)	4.6	832 (257)	1004 (314)	−172
Scene	786 (132)	957 (186)	−171	89.7 (9.74)	85.0 (14.6)	4.7	883 (152)	1180 (435)	−297

Figure 2 . Illustration of the stimuli for Experiment 4, showing a face that is intact and upright (left), with internal-upright features (middle), and with external-upright features (right).

Figure 2 presents three types of faces. The first face on the left is a frontal face. The face in the middle has the external features (e.g., head shape, hair) rotated 90° clockwise, with the internal features (e.g., eyes, mouth, nose) remaining upright. The final face on the right has the external features upright, whilst the internal features are rotated 90° clockwise.

Table 4 . Mean RTs (ms), Accuracy (%) and IES for Upright (U), External-upright (E), and Internal-upright Faces (I) in the Blank, Array and Scene Conditions in Experiment 4. Parentheses Show the Standard Deviation of the Mean.

	Upright	External-Upright	Internal-Upright	U-E	U-I	E-I
RTs
Blank	587 (103)	602 (86)	614 (110)	−15	−27	−12
Array	785 (107)	789 (111)	913 (141)	−4	−128	−124
Scene	828 (149)	838 (140)	945 (164)	−10	−117	−107
Accuracy
Blank	97.5 (4.61)	95.2 (8.71)	95.2 (8.71)	2.3	2.3	0
Array	96.3 (5.41)	94.4 (4.70)	91.4 (10.1)	1.9	4.9	3
Scene	92.0 (7.56)	95.4 (6.24)	86.6 (10.6)	−3.4	5.4	8.8
IES
Blank	602 (98)	639 (128)	658 (201)	−37	−56	−19
Array	815 (127)	892 (506)	1018 (506)	−77	−203	−126
Scene	899 (168)	874 (154)	1100 (238)	25	−201	−226

Figure 3 . Illustration of the stimuli for Experiment 5, showing an upright face on an original, unfiltered scene (top left), and scenes of high (top right), medium (bottom left) and low complexity (bottom right).

Read the detailed description of this figure

This Figure contains four examples of scenes with different levels of Voronoi tessellation. The top left scene is a photograph of a living room setting. Four armchairs are visible, along with a fireplace and a mantelpiece with vases and other display items. Bookshelves to the left and the right of the room are also visible as well as some paintings on the wall. The top right scene is the same image, except that a Voronoi filter has been applied so that the finer details of the image are not visible. The bottom left scene has a medium level of Voronoi filter so that the detail in the image is less complex and is a mosaic of colours. The bottom right scene has a low complexity Voronoi filter so that that image is made up of larger mosaic colour blocks with little detail. All scenes contain a frontal face in the bottom left quadrant of the image, unaffected by the Voronoi manipulations.

Table 5 . Mean RTs (ms), Accuracy (%) and IES for Upright (U) and Inverted (I) Faces at Original, High, Medium and Low Voronoi scenes in Experiment 5. Parentheses Show the Standard Deviation of the Mean.

	RTs			Accuracy			IES
	Upright	Inverted	U – I	Upright	Inverted	U-I	Upright	Inverted	U-I
Original	761 (181)	921 (241)	−160	90.0 (14.52)	81.5 (16.07)	8.5	871 (269)	1184 (450)	−313
High	730 (156)	733 (136)	−3	86.3 (13.15)	84.6 (12.60)	1.7	860 (197)	886 (216)	−26
Medium	620 (97)	668 (117)	−48	94.0 (8.00)	93.8 (6.74)	0.2	664 (116)	716 (139)	−52
Low	620 (106)	630 (139)	−10	97.1 (6.41)	96.9 (5.56)	0.2	640 (110)	652 (147)	−12