From:  Multimodal feature extraction and fusion for determining RGP lens specification base-curve through Pentacam images

 Results of map2 with a size of 88 × 88 × 3 without any preprocessing on maps and using the MNIST CAE architecture.

NumberMapsInput sizeFeaturesNo. of features PreprocessMSER2
1288 × 88 × 3img (map2)11,616no0.0160.5
2288 × 88 × 3area5no0.0140.15
32-digit10-0.0110.186
4-88 × 88 × 3digit + area10 + 5no0.0110.216

The bolded numbers in the table indicate the best results. CAE: convolutional autoencoder; MSE: mean squared error.