14p, Conclusion, better renderings than the previously-dominant approach of training deep convolutional networks to output discretized voxel representations.