<![CDATA[In the Digital Transformation era, many businesses use technology in the form of Deep Learning which is used to change the way business is run, one of the methods used is Emotion Recognition. Emotion Recognition itself is part of Computer Vision, and computer vision tasks are usually done using the CNN algorithm. Accuracy is important in Emotion Recognition where many studies use various methods, both Transfer and Hybrid learning to try to improve this aspect, so this research intends to design a Autoencoder + CNN + Attention that can be used for Emotion recognition, which is made by combining Encoder, CNN, and Attention Mechanisms. this model is circumspect by using FER2013 and compared to the CNN + Attention model which is shutting down in the same way. Even though the Autoencoder + CNN + Attention managed to get 64% Accuracy in Evaluate Test_Model compared to CNN + Attention which got 55%, it should be noted that adjustments still have to be treated because of the 43% sensitivity of testing on external data such as tuning, layer adjustments, and FER2013 data augmentation.]]>
Copyrights © 2023
