. 2021 May 12;41(19):4234–4252. doi: 10.1523/JNEUROSCI.1993-20.2021

Table 2.

The correlation coefficients of between-object Euclidean distances and within-object Euclidean distances with the rank order of brain regions or CNN layers in the four experiments

System name	Between-object Euclidean distance				Within-object Euclidean distance
System name	Position	Size	Image statistics	SF	Position	Size	Image statistics	SF
Brain - s (SE)	0.89 (0.03)	0.85 (0.03)	0.48 (0.20)	0.43 (0.15)	-0.88 (0.06)	-0.91 (0.03)	-0.87 (0.01)	-0.08 (0.17)
Brain - d (SE)	0.87 (0.04)	0.88 (0.02)	0.55 (0.25)	0.65 (0.11)	-0.80 (0.06)	-0.66 (0.11)	-0.90 (0.02)	0.10 (0.18)
Alexnet	0.47^*/^	0.50^*/^*	0.68	–0.74^*/^*	–0.58^*/^	–0.29^*/^	0.61^*/^*	–0.01
Cornet-S	0.97^/^	0.98^*/^*	0.96^*/^	0.61	–0.03^*/^*	0.64^*/^*	0.71^*/^*	0.89^*/^*
Densenet-201	0.92	0.96^/^	0.96^*/^	0.95^*/^*	0.19^*/^*	0.89^*/^*	0.92^*/^*	0.87^*/^
Googlenet	0.93	0.95^*/^	0.92^*/^†	–0.08^/^**	–0.60^*/^	0.13^*/^	0.53^*/^*	0.43^*
Inception_v3	0.92	0.96^/^	0.97^*/^	0.95^*/^*	0.28^*/^*	0.77^*/^*	0.85^*/^*	0.88^*/^*
Inception-resnet_v2	0.88	0.89	0.88^*	0.90^/^	–0.31^*/^	0.85^*/^*	0.78^*/^*	0.82^*/^
Mobilenet_v2	0.88	0.89	0.92^*/^†	0.81^*	–0.06^*/^*	0.66^*/^*	0.59^*/^*	0.77^*/^
Resnet-18	0.91	0.95^*/^	0.94^/^	0.85^**/^†	0.20^*/^*	0.77^*/^*	0.84^*/^*	0.86^*/^
Resnet-50	0.95^†/^†	0.97^*/^	0.96^*/^	0.71	0.23^*/^*	0.64^*/^*	0.68^*/^*	0.88^*/^*
Resnet-101	0.91	0.97^*/^	0.96^*/^	0.83^*	0.08^*/^*	0.59^*/^*	0.62^*/^*	0.84^*/^
Squeezenet	0.75^*/^	0.87	0.89^*	0.21^**	0.04^*/^*	0.58^*/^*	0.84^*/^*	0.84^*/^
Vgg-16	0.51^*/^	0.59^/^*	0.68	–0.15^/^*	–0.33^*/^	0.17^*/^	0.25^*/^*	0.00
Vgg-19	0.51^*/^	0.62^/^*	0.72	0.05^/^**	–0.24^*/^*	0.19^*/^	0.22^*/^*	0.24
Xception	0.81^*/^†	0.88	0.81	0.68	–0.17^*/^*	0.51^*/^*	0.09^*/^*	0.56^**/^†
Resnet-50-IN	0.95^†/^†	0.97^*/^	0.96^*/^	0.71	0.23^*/^*	0.64^*/^*	0.68^*/^*	0.88^*/^*
Resnet-50-SIN	0.95^*/^†	0.96^/^	0.95^*/^	0.93^*/^*	0.17^*/^*	0.73^*/^*	0.08^*/^*	0.89^*/^*
Resnet-50-SININ	0.91	0.96^*/^	0.94^/^	0.87^*/^	–0.08^*/^*	0.55^*/^*	0.47^*/^*	0.83^*/^
Resnet-50-SININ-IN	0.91	0.96^/^	0.95^*/^	0.88^*/^	–0.01^*/^*	0.62^*/^*	0.65^*/^*	0.89^*/^*

A positive/negative correlation indicates that Euclidean distances increase/decrease over the course of processing. The between-object Euclidean distance and the within-object Euclidean distance reflected the coding strength of object identity and nonidentity features, respectively. The brain results are the average correlation coefficients across the group of human participants, with Euclidean distances corrected by the reliability of each brain region using either the subtraction (brain - s) or division (brain - d) method (see Materials and Methods) before correlations are computed. To provide some quantification of the differences between the brain and the CNNs, the Fisher-transformed correlation coefficient of each CNN was tested against those across the group of human participants using two-tailed t tests. Significant differences found are marked with asterisks for the subtraction-corrected (left set) and division-corrected (right set) brain data.

†p < 0.1,

*p < 0.05,

**p < 0.01, and

***p < 0.001.

System name	Between-object Euclidean distance				Within-object Euclidean distance
System name	Position	Size	Image statistics	SF	Position	Size	Image statistics	SF
Brain - s (SE)	0.89 (0.03)	0.85 (0.03)	0.48 (0.20)	0.43 (0.15)	-0.88 (0.06)	-0.91 (0.03)	-0.87 (0.01)	-0.08 (0.17)
Brain - d (SE)	0.87 (0.04)	0.88 (0.02)	0.55 (0.25)	0.65 (0.11)	-0.80 (0.06)	-0.66 (0.11)	-0.90 (0.02)	0.10 (0.18)
Alexnet	0.47^*/^	0.50^*/^*	0.68	–0.74^*/^*	–0.58^*/^	–0.29^*/^	0.61^*/^*	–0.01
Cornet-S	0.97^/^	0.98^*/^*	0.96^*/^	0.61	–0.03^*/^*	0.64^*/^*	0.71^*/^*	0.89^*/^*
Densenet-201	0.92	0.96^/^	0.96^*/^	0.95^*/^*	0.19^*/^*	0.89^*/^*	0.92^*/^*	0.87^*/^
Googlenet	0.93	0.95^*/^	0.92^*/^†	–0.08^/^**	–0.60^*/^	0.13^*/^	0.53^*/^*	0.43^*
Inception_v3	0.92	0.96^/^	0.97^*/^	0.95^*/^*	0.28^*/^*	0.77^*/^*	0.85^*/^*	0.88^*/^*
Inception-resnet_v2	0.88	0.89	0.88^*	0.90^/^	–0.31^*/^	0.85^*/^*	0.78^*/^*	0.82^*/^
Mobilenet_v2	0.88	0.89	0.92^*/^†	0.81^*	–0.06^*/^*	0.66^*/^*	0.59^*/^*	0.77^*/^
Resnet-18	0.91	0.95^*/^	0.94^/^	0.85^**/^†	0.20^*/^*	0.77^*/^*	0.84^*/^*	0.86^*/^
Resnet-50	0.95^†/^†	0.97^*/^	0.96^*/^	0.71	0.23^*/^*	0.64^*/^*	0.68^*/^*	0.88^*/^*
Resnet-101	0.91	0.97^*/^	0.96^*/^	0.83^*	0.08^*/^*	0.59^*/^*	0.62^*/^*	0.84^*/^
Squeezenet	0.75^*/^	0.87	0.89^*	0.21^**	0.04^*/^*	0.58^*/^*	0.84^*/^*	0.84^*/^
Vgg-16	0.51^*/^	0.59^/^*	0.68	–0.15^/^*	–0.33^*/^	0.17^*/^	0.25^*/^*	0.00
Vgg-19	0.51^*/^	0.62^/^*	0.72	0.05^/^**	–0.24^*/^*	0.19^*/^	0.22^*/^*	0.24
Xception	0.81^*/^†	0.88	0.81	0.68	–0.17^*/^*	0.51^*/^*	0.09^*/^*	0.56^**/^†
Resnet-50-IN	0.95^†/^†	0.97^*/^	0.96^*/^	0.71	0.23^*/^*	0.64^*/^*	0.68^*/^*	0.88^*/^*
Resnet-50-SIN	0.95^*/^†	0.96^/^	0.95^*/^	0.93^*/^*	0.17^*/^*	0.73^*/^*	0.08^*/^*	0.89^*/^*
Resnet-50-SININ	0.91	0.96^*/^	0.94^/^	0.87^*/^	–0.08^*/^*	0.55^*/^*	0.47^*/^*	0.83^*/^
Resnet-50-SININ-IN	0.91	0.96^/^	0.95^*/^	0.88^*/^	–0.01^*/^*	0.62^*/^*	0.65^*/^*	0.89^*/^*