GET3D is a website offering a generative model of high quality 3D textured shapes learned from images. It uses adversarial losses defined on 2D images to generate 2D discriminators and a rasterization based differentiable renderer to obtain RGB images and silhouettes. Top 5 features:
– Generate 3D SDF and texture field via two latent codes
– Extract a 3D surface mesh from the SDF
– Query the texture field at surface points to get colors
– Train with adversarial losses defined on 2D images
– Utilize a rasterization based differentiable renderer to obtain RGB images and silhouettes
Top 5 Use Cases:
– Generate high quality 3D textured meshes ranging from cars chairs animals motorbikes and human characters to buildings
– Disentangle geometry and texture
– Smooth transition between different shapes for all categories
– Generate similar looking shapes with slight difference locally
– Produce meaningful view dependent lighting effects in a completely unsupervised manner.
