3D reconstruction of deformable objects with volume
Tipus de documentProjecte Final de Màster Oficial
Condicions d'accésAccés restringit per acord de confidencialitat
Reconstructing 3D models of deformable objects from monocular images or videos still remains a challenging--yet important--task in the field of Computer Vision. The aim of this work is to handle this problem and give a solution to both the model reconstruction and deformation tasks. We present our approach in two separate parts. First, we explain a method for approximating a three-dimensional model of an object from a single RGB image. Next, we propose an approach to sequentially obtain the deformed 3D shape of an object according to a new image in a monocular video, and the tracking of a few features. Technically, the two processes are independent, but it is clear how they naturally complement each other. The advantages--and limitations--of both methods are evaluated on real 3D objects.