Towards Practical Sketch-Based 3D Shape Generation: The Role of Professional Sketches

In this paper, for the first time, we investigate problem of generating 3D shapes from professional 2D sketches via deep learning. We target done by artists, as these are likely to contain more details than ones produced novices, and thus reconstruction such poses a higher demand on level detail in reconstructed models. This is importantly different previous work, where training testing was conducted either synthetic or novices. Novices often depict that physically unrealistic, while models trained with could not cope abstraction style found real sketches. To address problem, collected large-scale dataset sketches, each sketch paired reference shape, total 1,500 across 500 shapes. The available at http://sketchx.ai/downloads/. introduce two bespoke designs within adversarial network tackle imprecision human unique figure/ground ambiguity inherent sketch-based reconstruction. show existing generation methods designed images fail be naively applied our demonstrate effectiveness method both qualitatively quantitatively.