Projects

Light Field Synthesis

The aim of this project is to generate light field synthesis from single image. The deep learning frame work created takes only single image and generates multi plane image representation from which sub aperture views of light field are rendered. Disparity maps are also rendered from Multi Plane Image representation in the intermediate stage. Trained on open source dataset - Flowers dataset.

Polyp Segmentation from Gastro-Intestinal Endoscopy images

The aim of this project is to segment the polyp in gastro intestinal endoscopy images. Polyp is a abnormal portion in intestine looks like bulge like structure where it leads to cancer. I have used GAN based framework to train the dataset. I have designed a novel generator architecture for the faster inference. This network segmented polyps with high precision and accuracy. The dataset used is open source dataset which is Kvasir - SEG, Clinic DB dataset.

NLP Contribution Graph

In this project the contribution sentences are extracted from the research papers. Nowadays the number of publications are increasing every day. The research community should read many articles to get the gist on the problem which will be time taking. In our project we designed an AI model which extracts the contributing sentences and classify those sentences to some specific classes such that it indicates which section of the paper that contribution sentence belongs to. This gives the summary of the paper and saves the time in getting useful information from many bulk papers.

Emotion and Personality Detection from the English Speech Signal

In this project the emotion and personality are detected from the speech signal with an AI model. The correlation is established between the both emotion and personality. The open source data is used here.

AI Bot for Food ordering with RASA

This AI assistant helps in ordering food. It provides the menu. It records the order. It places the order. It handles the interruptions and chitchat well.

For the code repo click here.