It identified human emotion by partitioning photo into 4 segments with image processing function. To do this, I trained the recognition model using TensorFlow object with the data set around 400,000 faces from AffectNet. I compared between multiple algorithms include LeNet-5 CNN, Decision Tree, Random Forest, and AdaBoost and finally got around 70% accuracy. Finally, I extended the ML model to other use cases such as traffic lights detection with over 90% accuracy