Developed an Image Classification Deep Learning model with TensorFlow, Keras and OpenCV which is able to detect whether a person is happy or sad in any given image.

• Used TensorFlow metal for ARM-based mac devices.

• Loaded and cleaned image datasets.

• Splitted data into training, validation and test size.

• Used Adam optimisation.

• Visually displayed the model’s accuracy performance using Matplotlib’s plotting capabilities.

• Added 3 Convolutional and MaxPooling layers.

• Saved model in a ‘.h5’ file to enable model usability from outsiders.

• Created associated ‘.yaml’ file for creating a working conda virtual environment.

• Created a Basic Flask Application Front-End. Users can upload any picture of a person and output a result of the sentiment within the picture. It also gives an associated probability.

• Front-End is done with plain HTML.

• Implemented security features so that cookies sent and received from the browser are encrypted with a secret key.