If we wanted to map the colors of one model onto the shape of another model, would this mean that we would need both models to have the same topology? How would this work with models with different topologies?
S-Muddana
I don't think its a strict requirement that both models have the same topology. With appropriate correspondence establishment and parameterization techniques, color mapping can be achieved even for models with different topologies, although it may involve more complex and computationally intensive procedures.
brandonlouie
I think we may be able to employ something similar to the texture mapping portion of homework 1 in order to map the colors of one model onto another. If both topologies have the same number of triangles, then I'm willing to bet that we could apply barycentric coordinates to each triangle in model and use them as a mapping onto the other model. If they don't have the same number of triangles, I think we may be able collapse triangles similar to what's described in the later slides such that we can apply a similar mapping, although I believe this might cause for models to lose prominent features
ttalati
Is another use of having a common topology that for instance if the source and destination had access to the same basic, standard topology then if we wanted to communicate the specific geometry we can just send over the transform matrices information to convert into the specific geometry that we are communicating about? I am thinking this strategy may save some memory transfer/information transfer bandwidth, but I may be wrong. So is this another possible use for having common topology, specific geometry?
Alescontrela
I'm curious to see how this related to the SMPL human body models often used in Human pose estimation and generation. It seems like a lower-dimensional representation of the human body that ultimately maps to a bezier surface?
If we wanted to map the colors of one model onto the shape of another model, would this mean that we would need both models to have the same topology? How would this work with models with different topologies?
I don't think its a strict requirement that both models have the same topology. With appropriate correspondence establishment and parameterization techniques, color mapping can be achieved even for models with different topologies, although it may involve more complex and computationally intensive procedures.
I think we may be able to employ something similar to the texture mapping portion of homework 1 in order to map the colors of one model onto another. If both topologies have the same number of triangles, then I'm willing to bet that we could apply barycentric coordinates to each triangle in model and use them as a mapping onto the other model. If they don't have the same number of triangles, I think we may be able collapse triangles similar to what's described in the later slides such that we can apply a similar mapping, although I believe this might cause for models to lose prominent features
Is another use of having a common topology that for instance if the source and destination had access to the same basic, standard topology then if we wanted to communicate the specific geometry we can just send over the transform matrices information to convert into the specific geometry that we are communicating about? I am thinking this strategy may save some memory transfer/information transfer bandwidth, but I may be wrong. So is this another possible use for having common topology, specific geometry?
I'm curious to see how this related to the SMPL human body models often used in Human pose estimation and generation. It seems like a lower-dimensional representation of the human body that ultimately maps to a bezier surface?