SA-Optimized Hierarchical 3D Reconstruction NN

M.I. Fanany and I. Kumazawa (Japan)


3D reconstruction, neural network, simulated annealing


In this paper, we propose a new neural network (NN) design optimized by simulated annealing method for a robust 3D shape reconstruction. This NN deals with complicated task to reconstruct a complete rep resentation of a given object relying only on a limited number of views and erroneous depth maps of shaded images. The depth maps are obtained by Tsai-Shah shape-from-shading (SFS) algorithm. We investigate the effectiveness of this scheme through experiments of reconstructing a mannequin object from its images. The reconstruction result is evaluated for each view by comparing mean square error (MSE) of the depth map recovered by our reconstruction scheme and the depth map recovered by the SFS method alone, with a true depth map obtained by a 3D scanner. The ex perimental results show the SA optimization enable our reconstruction system to escape from local min ima. Hence, it gives more exact and stable results with small additional computation time.

Important Links:

Go Back

Rotating Call For Paper Image