The shallow neural network (SNN) model to predict NDVI from sRGB is loaded in the GitHub repository in the format “.h5” and will be accompanied by a Python demonstration function of its use which take the input from user (3 integer values from 0 to 255, corresponding to R-red, G-green and B-blue). These input data are converted to H (hue), S (saturation) and V (value) and all the six values (R, G, B, H, S, V) are collected in a data frame to be the input of the SNN model.
- Import of required libraries;
- Function (RGBtoNDVI) for ANN_saved_model.h5 application;
- Example of input take from user (R, G, B) and result prediction (NDVI).
- Number of samples: 4100
- Number of hidden layers: 1
- Number of neurons: 200
- % Training set: 75
- % Test set: 25
- r Training: 0.93
- r Test: 0.91
- TensorFlow
- Numpy
- Pandas
- colorsys
Lavinia Moscovini, Luciano Ortenzi, Federico Pallottino, Simone Figorilli, Simona Violino, Catello Pane, Valerio Capparella, Simone Vasta, Corrado Costa