stretcher seat back saddle body input sketch (a) (b) (c) (d) wing horizontal stablizer vertical stablizer engine propeller inferred structure labeling front frame wheel handle pedal chain fork back frame gas lift base rail stile arm Figure 1: Our method performs simultaneous part-level segmentation and labeling of input sketches (a), using database components and their interrelations. It first produces many local interpretations (b), which are optimized into a global interpretation (c) that fits the input sketches as well as forming plausible structures, with which the input sketches can be appropriately labeled (d).
AbstractWe present a data-driven approach to derive part-level segmentation and labeling of free-hand sketches, which depict single objects with multiple parts. Our method performs segmentation and labeling simultaneously, by inferring a structure that best fits the input sketch, through selecting and connecting 3D components in the database. The problem is formulated using Mixed Integer Programming, which optimizes over both the local fitness of the selected components and the global plausibility of the connected structure. Evaluations show that our algorithm is significantly better than the straightforward approaches based on direct retrieval or part assembly, and can effectively handle challenging variations in the sketch.