Achievements

Storms End: AI enabled solution to visualize and predict Geomagnetic Storms from NASAs' "DISCOVR" spacecraft

Mainul Islam Labib (Team Leader)

Md. Johir Raihan

Wali Ullah

Sadia Afrin

Md. Rashed Jaowad Khan

Competition NASA Space Apps Challenge 2023

We are proposing "Storms End"; an AI enabled solution to visualize and predict geomagnetic storm. Our solution includes a Neural Network consisting of Spatial and Temporal modules to correctly capture the variability in magnetic readings made from "DSCOVR". The model will be able to take previous readings as input and will output the probability for a geomagnetic storm on a spatio-temporal scale. "Storms End" will also have visualization features where users will be able to see geomagnetic patterns in a 3D environment and how each area on earth is getting affected by geomagnetic storm.

Autistic Kids Companion: An Automatic Exercise Assessment and Companion for Autistic Kids and Motor Dysfunction Patients Using ML.

Md. Johir Raihan

Competition Mobile App Game & Job Festival, ICT, Bangladesh 2023

The app titled “Autistic Kids Companion” uses a Mediapipe-based pose landmark algorithm to extract landmarks which are given to a Machine Learning (ML) model to predict the score of the performed exercise and analyze their body movement. The app will also provide which body part they should focus more on to accurately do the exercise. I have also provided an idea to use animated characters in place of their body so that it is more entertaining for them.

Assessing The Rehabilitation Exercise Using GWO Optimized CNN-SE Network and Interpretation of The Model Using SHAP.

Md. Johir Raihan

Md Atiqur Rahman Ahad

Abdullah-Al Nahid

Summer Symposium IEEE CS BDC Summer Symposium 2023, IEEE Computer Society Bangladesh Chapter 2023

The rehabilitation exercise assessment evaluates a person's physical capabilities and guides them on how the exercise should be done to improve their physical ability. Even after automating the process of assessing the rehabilitation exercise score using modern Machine Learning (ML) or Deep learning (DL) techniques, their black-box nature holds them back from use in the real world. Able to interpret these models and finding out how these models are using the human body pose to predict an exercise score may give a better understanding of both the model and the person's data that have been used