Customizing Deep Learning Based Facial Expression Recognition Model

Project Discription

In the world of artificial intelligence the new milestones can be set and achieved through facial expression recognition. By observing the human's emotion information, machines can provide tailored facilities .It can be applicable for many observations, such as virtual reality, personalized recommendations, and so on. This field having the potential for researchers to develop such a technique which should be able extract features correctly without distracting from head pose or occluding conditions. So to recognize the expression in uncontrolled environment with higher accuracy and in less time must be the motive.

A new approach was incorporate for identifying the facial sentiment of the input data. In this work input images were preprocess by convolution steps where input image was passed through various layers of processing. Once that processing complete then the training of neural network was done by softmax function. In this whole process of sentiment leering from input images, execution time was quite high. At the same time due to use of lots of filter in various layers of convolution neural network accuracy of detection was also less.

This work has classified kind of information, obtained from input facial image with tag of sentiment information. Entire work is clarified in this segment of the work and block diagram of fig. demonstrates all means of the work.

Project Sample Image