Fig. 1. The structure of GFCN model

Fig. 2. The spectra after preprocess
(a): Tecator dataset; (b): IDRC2018 dataset

Fig. 3. Training loss and test loss of ANN model
(a): FCN; (b): GFCN

Fig. 4. Influences of the number of groups on the predicting effect of GFCN model

Fig. 5. Prediction results of three models
(a): Fat; (b): Moisture; (c): Protein; (d): IDRC2018

Fig. 6. The influence of the size of training set on the predictings effect of three models

Fig. 7. Contribution ratio of each wavelength of FCN and GFCN models for predicting fat component