Intelligent Classrooms

Finalist at Singapore-India Hackathon (Top 5) with cash prize of $2,000 ; Project Idea given special mention by India PM Modi at awards ceremony

Built a end-to-end architectural system that incorporates human pose estimator, emotion recognition and head gaze deep learning model into a customised neural network

Architecture

Teachl.AI employes a suite of 3 computer vision models that enables a holistic understanding of a student

1. Emotion Recognition
2. Human Pose Estimation
3. Gaze Detection

These outputs are harnessed as inputs into a customized neural network that generates prediction of engagement levels of the student.

The engagement level is then displayed in a classroom heatmap to show the dynamic changes in the classroom engagement. Engagement levels are also plotted onto a dynamic chart to monitor the lesson across the lessons and help teachers to find the optimal time to call for breaks.

Built With

• c
• css
• html
• javascript
• jupyter-notebook
• perl
• powershell
• python
• roff
• sass
• scss
• shell