Research

Image based 3D modeling

[ tree modeling ]

In this project we developed a novel approach for reconstructing 3D models of tree-like objects by using a sparse set of viewpoints. Previous approaches either requires a dense set of images or relies on computational tomography for dense point cloud registration. In contrast, our approach focuses on faithfully recovering real instead of realistically-looking tree geometry from a sparse set of images.
Our solution actively integrates 2D and 3D tree topology as shape priors in the reconstruction process and it is able to accurately recover complex tree geometry with minimal user interventions. Image Modeling Project Webpage



[ root modeling ]

In this paper we present an image-based system for recovering complex 3D model of real roots. Our method is able to recover accurate 3D root geometry. Compared to existing reconstruction approaches, our system can accurately recover complex topology and geometry of 3D roots using only a few set of images. Our system can generate fine geometry of roots in few minutes with minimal user intervention, making it more suitable for large scale applications. Image Modeling Project Webpage

Biomedical Figure and Text Mining

[ Project Webpage ]


[ Automatic Extraction of Figure-Caption Pairs from PDF Files ]
In this paper we present a novel approach for automatically extracting figure-caption pairs from Biomedical publications in PDF format.