Meta publishes "Digital Twin Catalog" with 2,400 photorealistic 3D models
Take a few photos of an object and get an accurate 3D replica for XR or e-commerce applications: Meta's Digital Twin Catalogue will help develop such 3D reconstruction techniques.
Reality Labs Research, Meta's VR and AR research division , has released the Digital Twin Catalogue (DTC). According to Meta, this is the world's largest and highest quality dataset of 3D object models.
The DTC contains over 2,400 highly detailed 3D object models that are photo-realistic and accurate to the sub-millimetre scale of their physical counterparts. The objects are everyday items such as toys and kitchen utensils.
Researchers can now use the Digital Twin Catalog to train, optimize and evaluate AI-based 3D reconstruction techniques.
Meta also announced a partnership with Shopify, a leading global e-commerce company. Over the next few months, the partners will provide more than 7,000 digital twins as part of an academic competition to further improve the reconstruction algorithms.
Digital twins are photorealistic 3D object models that are indistinguishable from their physical counterparts. The vast majority of digital twins in e-commerce and immersive applications today are painstakingly created by hand by teams of technical artists. In some cases, the cost can be as high as the production cost of the physical object.
To simplify 3D reconstruction, Meta has been investing in the development of techniques for many years.
As with all machine learning-based methods, data will be the key to further advances in 3D reconstruction, Meta writes on its Meta AI blog. 3D object datasets have grown in recent years, but still lack geometric and material fidelity. With the DTC, Meta aims to help make 3D reconstruction more robust and efficient.
Note: Links to online stores in articles can be so-called affiliate links. If you buy through this link, MIXED receives a commission from the provider. For you the price does not change.
