Unsupervised bas-relief generation with angle defect

The generation of bas-relief based on 3D models has always been a research hotspot in computer graphics. Traditional algorithms are computationally complex. The latest methods based on deep learning bring new ideas to the problem but are rely on supervised training, which requires a large number of samples to be constructed manually. In this paper, an unsupervised bas-relief generation method is proposed. In this method, we take the depth field of point cloud as input, and reconstruct the bas-relief using a Convolutional Neural Networks. During the process, the smallest change of angle defect is selected as the objective of optimization, thus, avoids the constructing of the "ground truth". Experiments show that the method in this paper is simple and efficient, and can generate bas-relief models with good saturation and detail feature.