Page 89 - Read Online
P. 89

Page 83                               Bah et al. Intell Robot 2022;2(1):72­88  I http://dx.doi.org/10.20517/ir.2021.16
































               Figure 5. Confusion matrix for the FER showing the prediction accuracy of the emotional expressions using the proposed architecture with
               residual blocks (left) and the basic architecture (right).







































                      Figure 6. The accuracy and the loss of the training and validation data of FERGIT on the basic model over 100 epochs.

               To validate our network’s capability of fast generalization and giving the best accuracy, we also trained it on
               the CK+ database [43] . The CK+ database is relatively small, with 981 samples well partitioned with seven
               classifications of emotions: Angry, Disgust, Fear, Happy, Sadness, Surprise, and Contempt [43] . Using dropout
               layers helped the model train on a very small dataset (see Figure 8). The model achieved competitive results on
   84   85   86   87   88   89   90   91   92   93   94