This was my final project for the Bachelor's Thesis. It is a system that can monitor the emotions of the people who are taking online exams, through voice and video. For full information about the whole system, please check out the Documentation which contains my Bachelor's Thesis.

There are two APIs and one client. The AI API can be accessed only through requests to the CRUD API. That offers more scalability as if by instance somebody wants to implement their own AI in this project, they can connect the CRUD API requests to their own API

The system monitors people's emotions while taking online tests. While solving the test, a thread is open and periodically it takes capture of the web camera, sending them to the CRUD API through base64 format, ending up to the AI API and predicting the emotion for that capture. The same flow is applied for the speech channel, with the only difference that the user will need to press a button in order to register their answer.

For facial detection we used FER2013.

For speech detection we used RAVDESS.

For more information about the two datasets, please read the Documentation !

The two models are public on Kaggle. For those who want to use my system, you will need to retrain the models from scratch and include them in the AI API. Here are the notebooks for the facial model and the speech model:

https://www.kaggle.com/code/razvanispas/ravdess-densenet-fine-tuned

- FastAPI
- PostgreSQL 
- SQLModel
- Vue.js 2.0
- Vuetify 
- TensorFlow
- CV2 

We recommend a conda environment, but if you prefer a pip environment, you can install everything required for the APIs by using the requirements file:

pip install requirements.txt

For the client, Vue.js 2.0 with Vuetify and Chart.js will be required. It is expected from you to configure the Vue.js environment such that the components provided in the Client folder to work properly.

This was my project for my Bachelor's thesis. Anyone who will continue my work for their thesis will be accused of plagiarism.