Utilizing deep learning framework using Tensorflow and Keras library to create facial emotion recognition with transfer learning technique based on Convolution Network (CNN),VGG16 and Resnet50 architecture. Advance features including using tensorboard HPARAM to tune hyperparameter and created input pipeline and analysis on pipeline bottleneck in order for the framework to be compatible with big data.

Convolution Network Accuracy

Convolution Network Loss

VGG16 Accuracy

VGG16 Loss

Resnet50 Accuracy

Resnet50 Loss

Hyperparameter tuning for VGG

Hyperparameter tuning for Resnet

Overall Result for the experiment