When the curvature of the nearby triangles is small, it means a group of triangles shares the similar normal direction, therefore I believe that a weighted normal direction computed out of them can a simplification for this area. But what is the most suitable weight assignment? Could it be the area value of triangles? (I think the weight values can be normalized by dividing the sum of all area values)
aravind00r
There are definitely many good algorithmic down-sampling methods out there, but in my experience, many 3D artists use a slightly more manual approach (https://www.youtube.com/watch?v=BEwEWKOH5ws&ab_channel=FlippedNormals). This process is called retopology. Often artists will sculpt very vertex dense models with unnecessarily high detail, and then they will use this process to remove any unnecessary details in regions where it is not needed (naturally flat or plain sections of the model). This way you are not uniformly down-sampling the model, and you can steer what parts of the model you are willing to lose resolution in.
When the curvature of the nearby triangles is small, it means a group of triangles shares the similar normal direction, therefore I believe that a weighted normal direction computed out of them can a simplification for this area. But what is the most suitable weight assignment? Could it be the area value of triangles? (I think the weight values can be normalized by dividing the sum of all area values)
There are definitely many good algorithmic down-sampling methods out there, but in my experience, many 3D artists use a slightly more manual approach (https://www.youtube.com/watch?v=BEwEWKOH5ws&ab_channel=FlippedNormals). This process is called retopology. Often artists will sculpt very vertex dense models with unnecessarily high detail, and then they will use this process to remove any unnecessary details in regions where it is not needed (naturally flat or plain sections of the model). This way you are not uniformly down-sampling the model, and you can steer what parts of the model you are willing to lose resolution in.