Academia.eduAcademia.edu

Figure 8 – uploaded by Nico Pietroni

See full PDFdownloadDownload figure
Fig. 7. The effect of energy terms in the segmentation results: (a) a detail of the final segmentation on a model, resulting in just two segments, and smooth boundaries; (b) the segmentation if plain geodesics were used in the data cost, instead of tubeness-weighted geodesics: the segment which wraps around the model arm (circled in red) would lead to a non-extractable mold piece; (c) the effect of neglecting the smoothing cost, namely poor boundary quality and isolated fragments near the boundary (circled in red); (d) the result if no label costs were imposed, with too many pieces to be useful in practice.

Figure 7 The effect of energy terms in the segmentation results: (a) a detail of the final segmentation on a model, resulting in just two segments, and smooth boundaries; (b) the segmentation if plain geodesics were used in the data cost, instead of tubeness-weighted geodesics: the segment which wraps around the model arm (circled in red) would lead to a non-extractable mold piece; (c) the effect of neglecting the smoothing cost, namely poor boundary quality and isolated fragments near the boundary (circled in red); (d) the result if no label costs were imposed, with too many pieces to be useful in practice.

Related Figures (21)

Fig. 3. The k = 650, uniformly sampled, candidate parting directions. 36:4 « Thomas Alderighi, Luigi Malomo, Daniela Giorgi, Nico Pietroni, Bernd Bickel, and Paolo Cignoni
Fig. 3. The k = 650, uniformly sampled, candidate parting directions. 36:4 « Thomas Alderighi, Luigi Malomo, Daniela Giorgi, Nico Pietroni, Bernd Bickel, and Paolo Cignoni
Fig. 4. (a) Visible regions from a given direction are depicted in red. (b) The corresponding moldability field takes a zero value over visible faces, and is interpolated over the rest of the surface through a shape-aware geodesic flow.
Fig. 4. (a) Visible regions from a given direction are depicted in red. (b) The corresponding moldability field takes a zero value over visible faces, and is interpolated over the rest of the surface through a shape-aware geodesic flow.
Fig. 5. Visibility computation scheme.
Fig. 5. Visibility computation scheme.
Fig. 6. (a) Tubeness computation at a vertex v, by tracking the topology of the surface region defined by the points having geodesic distance from v less than a varying radius. On tube-like features, the surface region changes topology for a small radius R; (b) Tubeness values on two example surfaces.
Fig. 6. (a) Tubeness computation at a vertex v, by tracking the topology of the surface region defined by the points having geodesic distance from v less than a varying radius. On tube-like features, the surface region changes topology for a small radius R; (b) Tubeness values on two example surfaces.
Fig. 8. Segmenting via a two-stage strategy: (a) anisotropic clustering of mesh faces; (b) segmenting the clustered mesh to get optimal parting directi- ons; (c) segmenting the original mesh with respect to the optimal directions. 4.4 Two-stage optimization approach
Fig. 8. Segmenting via a two-stage strategy: (a) anisotropic clustering of mesh faces; (b) segmenting the clustered mesh to get optimal parting directi- ons; (c) segmenting the original mesh with respect to the optimal directions. 4.4 Two-stage optimization approach
Fig. 9. Dealing with positive-genus models: (a) computation of a shortest basis of tunnel loops; (b) topological simplification by inserting membranes; (c) segmentation of the remeshed model; (d) keeping membranes in the metamold to define proper cuts in the mold volume.
Fig. 9. Dealing with positive-genus models: (a) computation of a shortest basis of tunnel loops; (b) topological simplification by inserting membranes; (c) segmentation of the remeshed model; (d) keeping membranes in the metamold to define proper cuts in the mold volume.
Fig. 10. Modeling parting surfaces: (a) the oriented sampling on the boundary of a segment: the normal to each sample (green) is the cross product betwe the normal to the original surface (blue) and the direction along the segment boundary (purple); (b) the parting surfaces for the armadillo model; (c) intert membranes and parting surfaces for an object with high genus; (d) parting surfaces for a model segmented into three pieces.
Fig. 10. Modeling parting surfaces: (a) the oriented sampling on the boundary of a segment: the normal to each sample (green) is the cross product betwe the normal to the original surface (blue) and the direction along the segment boundary (purple); (b) the parting surfaces for the armadillo model; (c) intert membranes and parting surfaces for an object with high genus; (d) parting surfaces for a model segmented into three pieces.
Fig. 11. A visualization of the trapped air problem for both silicone and resin casting. During silicone casting, the mold volume (green area) could trap air bubbles around downward maxima (blue dot). During resin casting, the model cavity (pink area) could trap air bubbles in correspondence of upward maxima (yellow dots).
Fig. 11. A visualization of the trapped air problem for both silicone and resin casting. During silicone casting, the mold volume (green area) could trap air bubbles around downward maxima (blue dot). During resin casting, the model cavity (pink area) could trap air bubbles in correspondence of upward maxima (yellow dots).
Fig. 12. (a) Details on the metamold geometry: 3D printed pipes (in blue) anchored to pegs, and sealing dams closing a loop around the object on the parting surface; (b) the 3D printed pipes create air vents in the mold volume, to let air escape while casting. Given the final pouring directions, we automatically add the final details over the metamolds geometry. Small pegs are placed in correspondence of each local maximum, defining the anchor points for 3D printed pipes that will be attached to form the air vents in the mold volume. We take the global maximum as the anchor point for a larger vent, from which resin will be poured. Sealing dams are added as a plug and slot structure surrounding the object in a closed loop over the parting surface to secure the sealing between different mold pieces (Figures 12.a and .b). Also, to ensure a perfect interlock of the mold pieces, registration pegs and holes are added on the parting surface in correspondence of topological membranes shared among different mold pieces.
Fig. 12. (a) Details on the metamold geometry: 3D printed pipes (in blue) anchored to pegs, and sealing dams closing a loop around the object on the parting surface; (b) the 3D printed pipes create air vents in the mold volume, to let air escape while casting. Given the final pouring directions, we automatically add the final details over the metamolds geometry. Small pegs are placed in correspondence of each local maximum, defining the anchor points for 3D printed pipes that will be attached to form the air vents in the mold volume. We take the global maximum as the anchor point for a larger vent, from which resin will be poured. Sealing dams are added as a plug and slot structure surrounding the object in a closed loop over the parting surface to secure the sealing between different mold pieces (Figures 12.a and .b). Also, to ensure a perfect interlock of the mold pieces, registration pegs and holes are added on the parting surface in correspondence of topological membranes shared among different mold pieces.
Table 1. Timings for deriving the optimal segmentation.
Table 1. Timings for deriving the optimal segmentation.
Fig. 13. 3D printed walls used for the silicone pouring procedure.
Fig. 13. 3D printed walls used for the silicone pouring procedure.
Fig. 14. Resulting segmentation for basic shapes. Red arrows represent the extraction direction chosen for each segment.
Fig. 14. Resulting segmentation for basic shapes. Red arrows represent the extraction direction chosen for each segment.
Fig. 15. Some simple models cast with our technique. Left: the optimized segmentation; right: the fabricated metamolds and the cast object.
Fig. 15. Some simple models cast with our technique. Left: the optimized segmentation; right: the fabricated metamolds and the cast object.
Fig. 16. Some more complex models cast with our technique. Left: optimized segmentation; right: fabricated metamolds, molds and cast object.
Fig. 16. Some more complex models cast with our technique. Left: optimized segmentation; right: fabricated metamolds, molds and cast object.
Fig. 20. A failure case: a model with extremely long thin features may require additional cuts that cannot be detected by our technique. Fig. 19. Successful casting of objects which could not be fabricated with the approach in [Herholz et al. 2015]. The heptoroid (third row) is a failure case alsc for FlexMolds [Malomo et al. 2016], which derives an extremely complex cut layout (left), whereas our approach requires just a two-piece mold (right).
Fig. 20. A failure case: a model with extremely long thin features may require additional cuts that cannot be detected by our technique. Fig. 19. Successful casting of objects which could not be fabricated with the approach in [Herholz et al. 2015]. The heptoroid (third row) is a failure case alsc for FlexMolds [Malomo et al. 2016], which derives an extremely complex cut layout (left), whereas our approach requires just a two-piece mold (right).
Fig. 21. An Ice-Age bunny.
Fig. 21. An Ice-Age bunny.
Related topics:
Information SystemsComputer Science
580 California St., Suite 400
San Francisco, CA, 94104
© 2025 Academia. All rights reserved