CNN, Streamlit, Heroku Deployment, WebCam Feed
The Indian education landscape has been undergoing rapid changes for the past 10 years owing to the advancement of web-based learning services, specifically, E-Learning platforms.
In a physical classroom during a lecture, the teacher can see the faces and assess the emotion of the class and tune their lecture accordingly, whether he is going fast or slow. He can identify students who need special attention. Digital classrooms are conducted via a video telephony software program (ex- Zoom) where it’s not possible Convene to see all students and access the mood. Because of this drawback, students are not focusing on content due to a lack of surveillance. Digital platforms have limitations in terms of physical surveillance but it comes with the power of data and machines which can work for you. Its data can be analyzed using deep learning algorithms which not only solve the surveillance issue but also remove the human bias from the system. To construct and face emotion recognition model for Live Class Monitoring System.
Steps involved:
- Exploratory Data Analysis
- Data Image generator
- CNN Model
- Training the model
- Model Evaluation
- Model Deployment
Following conclusions can be made from project:
- The CNN model gave us training accuracy of 70 % and validation accuracy of 63 %.
- The application is able to detect face location and predict the right expression while checking it on a local webcam.
- A front-end model was successfully created using streamlit and run on a local webserver.
- Successfully deployed streamlit web app on Heroku and streamlit share that runs on a web server.
- Through this model teachers can understand the students' perception during online classes and change the way of teaching if needed by understanding the students’ motive.