DCod.X

Analyze real-time online behavioral activities of staff members and students, identify their state of minds and emotions

About the Team:

Problem statement:

> The Problem being solved here is to make an application that will help the teachers/professors to analyze student's behavioral aspects through the detection of their facial expressions, like, attentive, lazy, sad, happy, surprised etc.

> It will also detect the scamming video feeds that are adopted by various students in order to skip from their live lectures.

Solution:

> To make a web or software-based application that could be easily integrated with any online classroom or office meetings platforms or software like ZOOM, GOOGLE MEET, Microsoft Teams etc.

> It will be fed with some facial models in the behind to monitor the student’s or staff member’s facial expressions to recognize their state of mind and video delays.

> All the monitoring will be done through an online decentralized server or computer.

> Live facial expressions will be translated into text and will be displayed to the professor or the teachers. By evaluating the emotional state, there is an attempt to overcome the barrier between man and non-emotional machine.

Language/Tools:

> We used Python programming language as backend language, for frontend/GUI we used flask and trinket and conversion of visual images into text we used Matplotlib library.

> We used Python due to its vast machine learning and deep learning libraries, as it will help us train our application on real-time data set models and deduce results accordingly.

> Other major reason of selecting Python is due to its versatile nature, like we can make GUI based application or a website from it, we can also design and make different API’s (like flask API that will be used in final development of the project) with the help of python.

Algorithm:

> We used Stochastic Gradient Descent image recognition algorithm for the detection of images in this project/application.

> The reason behind using this algorithm is that its mini batch stochastic gradients Descent and variants, thereof have become standard for large scale empirical risk minimization like the training of neural networks.

> These methods are usually used with a constant batch size using a simple empirical inspection.

> When using a constant batch size, stability and convergence is thus often enforced by means of a decreasing learning rate schedule.

UI/UX Screens:

> Login Screen:

> Evaluation/Options Screen:

> Results Screen:

BlockChain Element:

> We introduced decetralized image recognition API in this application/prototype/project so that the machine will be trained but the data will be erased after the end of every session.

> We created a complete login panel on which only registered member can have access to the application.