Abstract:

Charts are commonly used to understand data in digital documents such as websites, digital papers. It also helps to understand the relationship between parts of data. But it is hard to understand for blind people because charts rely on visuals only. This workgives you accessibility to understand pie charts for blind people. Given a 2D-chart bitmap image as an input, it separates only pie chart images by applying a deep learning-based object detection model. Then it extracts the underlying data using different Computer vision techniques (like OCR, Circle detection, color detection, pixel detection) to extract data from a chart image and use this information as a HashMap to use later. We generate images using three common Python libraries and apply that dataset to our algorithm. There are challenges i.e., text written in pie chart is hard to read and obstacle for extracting data in percentage. In order to complete this, we are proposing a colord-pixel-based solution to overcome these problems.

Zoom link: https://lums-edu-pk.zoom.us/j/92971501534?pwd=UjJkUWxib0l6UmlYdHJxNEhwU1c2Zz09

Meeting ID: 929 7150 1534

Passcode: 979645