Figure 1.

Prediction error (predicted age minus true age) for the same 25 images with various resolutions. Images were passed through the developed neural network, with kernels trained for 299 x 299 pixels resolution.