Interactive Machine Learning

Current machine learning based systems mostly work in inference mode, which implies that one gives an input to the system, and gets an output. Thereafter, the only choice is to accept or reject the output. We have developed a new algorithm, which allows “human in the loop” design. Here an ML system produces an output (possibility highly erroneous), a doctor sees it and provide quick and limited corrections. The ML model takes the feedback into account, retrains itself in real time, and produces a new output. The doctor sees the new output and corrects again. The iterations continue until a doctor is mostly satisfied with the output.

Ayu - Characterization of Healthy Ageing from Neuroimaging Data with Deep Learning and rsfMRI

In this work, we propose an age prediction pipeline Ayu which consists of data preprocessing, feature selection and an attention-based model for deep learning architecture for brain age assessment on a dataset (N = 638, age-range 20-88) comprising rsfMRI images from the Cambridge Centre for Ageing and Neuroscience (Cam-CAN) repository of a healthy population.

Covid Cough Detector

In this work, I worked on a Covid-19 Cough Detector using MFCCs and Chroma based features from 10 seconds cough audio recording of the person and applying a CNN based Architecture to classify COVID-19 positive patient or not.
This work was done on IISc ”Co-Swara” Dataset.