A new convolutional neural network (CNN) has been developed to address the problem by a research team led by Dr. Jia Tian from Jilin Engineering Normal University in China. His team concentrated on getting an optimal balance between the training speed, memory usage, and recognition accuracy of the model.<\/p>\n
“The model we developed is particularly effective for facial expression recognition when using small sample datasets. The next step in our research is to further optimize the model’s architecture and achieve an even better classification performance,”<\/p><\/blockquote>\n
Tian said.<\/p>\n
58,000 Parameters<\/h2>\n
Structure of the proposed neural network with depthwise separable convolutions and pre-activated residual units. Credit: J. Tian et al.<\/p><\/div>\n
Among the differences between typical CNN models and the one envisioned by the researchers was the use of depth-wise separable convolutions. Convolutions are the core operation executed at each layer of a convolutional neural network.<\/p>\n
Where a depth-wise separable convolution differs from the standard type is that it processes different channels (such as RGB) of the input image independently and combines the results at the end.<\/p>\n
By blending this type of convolution with a technique called pre-activated residual blocks<\/a>, the proposed model was able to process input facial expressions in a coarse-to-fine manner. In this way, the team greatly reduced the computational cost and the necessary number of parameters to be learned by the system for accurate classification.<\/p>\n
“We managed to obtain a model with good generalization ability with as little as 58,000 parameters,”<\/p><\/blockquote>\n
Tian said.<\/p>\n
72 Percent Accurate<\/h2>\n
The researchers tested their new model through a comparison of its facial expression recognition performance with that of other reported models in a classroom setting. They trained and tested all models using a popular dataset called the Extended Cohn-Kanade dataset<\/a>, which contains over 35,000 labeled images of faces expressing common emotions.<\/p>\n
The results were encouraging, with the model developed by Tian’s team exhibiting the highest accuracy (72.4%) with the least number of parameters.<\/p>\n
![\"Structure](\"https:\/\/techfragments.com\/wp-content\/uploads\/2022\/07\/convolutional-net-facial.jpg\")