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

 Results of the proposed CAE architecture and the effect of the preprocessing module on images to extract features and calculate color areas.

NumberMapsInput sizeFeaturesNo. of featuresPreprocessMSER2
11, 2, 3, 481 × 81 × 12img (all maps)12 (1 × 1 × 12)Yes0.0180.267
2281 × 81 × 3area7Yes0.024 0.023
No0.0120.262
3--digit10-0.0110.315
4281 × 81 × 3area + digit19 (7 + 10)Yes0.007 0.699
No0.0110.318
51, 2, 3, 481 × 81 × 12area + img19 (7 + 12)Yes0.017 0.273
No0.020.167
61, 2, 3, 481 × 81 × 12digit + img22 (10 + 12)Yes0.0080.669
71, 2, 3, 481 × 81 × 12area + digit + img29 (7 + 10 + 12)Yes0.0090.639
No0.0080.646

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