Model
DeepSDF
A neural implicit 3D shape representation method based on signed distance functions. It learns per-shape latent codes to achieve cross-shape generalization, enabling rapid inference for unseen shapes via latent code optimization.
Timeline (last 90 days)
Oct 5
Neural representations trained by overfitting to a single shape are called overfitting-based or per-shape fitting methods, while methods with generalization capability are called feed-forward methods (e.g., DeepSDF, OccNet).
Unverified50%