Table 3.

Validation losses and accuracies of deep-convolutional neural networks. Each architecture was transfer learned with a dataset of blastocysts and non-blastocysts imaged at 113 hpi. The error values reported are standard errors of mean.

Architectures	Validation losses	Validation accuracies (%)
Xception	0.86 ± 0.003	63.53 ± 0.631
ResNET-50	0.88 ± 0.002	59.97 ± 1.08
Inception v3	0.91 ± 0.01	61.57 ± 0.689
NASNetLarge	1.3 ± 0.004	45.75 ± 1.052
Multilayer CNN	1.14 ± 0.009	49.17 ± 1.108
ResNeXt-101	0.95 ± 0.036	58.17 ± 1.2
ResNeXt-50	0.99 ± 0.029	60.07 ± 2.076
Inception ResNET-V2	0.87 ± 0.005	62.09 ± 1.342