A Survey of Learning-Based Mesh Processing Techniques For Mesh Simplification

With a new industrial revolution on the horizon, companies are beginning to increasingly adopt advanced technologies into their production environment, including the scanning and processing of dense 3D meshes of products. In the field of 3D mesh processing, machine learning has emerged as a tool to assist in the simplification of high-poly meshes. In this context, a survey was conducted with the purpose of determining the current state of the literature and work produced in the fields of machine learning and computer graphics with regards to mesh processing and simplification. In this survey, we compile a list of developed frameworks, briefly summarize how they work and their contribution to the scientific knowledge available at the time, and compare each framework in terms of scope, use cases, and side-by-side performance. In the end, we see clearly how researchers have been progressively, over the last five years, expanding on the previous works of their peers, achieving superior mesh simplification accuracy and enabling the encoding of higher resolution meshes.