Machine Learning
Education Technology

Segmentation of sparse images of charts for extracting components and locating salient trends depicted in the charts

Project Principal Investigator(s): Jaya Sreevalsan Nair

A research project on segmenting an image of a chart, with an initial focus on a bar chart, to localize text, axes, graph area, and legends as components. Once this is done using state-of-the-art methods, such as CNN, to accommodate the sparsity of such images, we focus on feature mapping and ranking by identifying persistent features, which effectively catch the viewer’s eye. We also use validation from high school students’ interpretation of charts to identify appropriate underlying computational models to depict charts.

This project will entail computer vision, image processing, and machine learning for identifying components in the images of charts. This will be followed by extraction of information from the charts.This will entail feature tagging in large database of images of bar charts