The problem Sentimentaly solves

We aim to tackle

• It can be used on films to give an unbiased analysis of emotions conveyed through the movie
• Closed market studies can use this as a tool to ascertain user response to their products with better data than using conventional methods like survey forms.
• Public speakers and Presenters can use it to monitor crowd responses and get real time feedback from their audiences to adjust their presentation techniques accordingly.
• Actors and people preparing for interviews or speeches can use it as a mirror to get feedback
• User while uploading pictures to social media can check whether each face on the image shows the right emotion

Currently it has been developed as an android prototype and can be expanded as an API and for edge devices like raspberry pi

Challenges we ran into

Deep Learning

• Data Collection
• Data Cleaning
• Optimising Model for edge devices
• Running out of training time on cloud-GPU

Android

• Deploying TFlite model
• Integration of Mobile Vision
• Accessing GPU on available android device