Gaussian Masked Autoencoders
This addresses the problem of limited spatial understanding in self-supervised learning for computer vision researchers, offering a novel integration of Gaussian primitives into representation learning.
The paper tackles the lack of explicit spatial awareness in Masked Autoencoders (MAE) by introducing Gaussian Masked Autoencoder (GMAE), which learns semantic abstractions and spatial understanding jointly through a 3D Gaussian-based representation and splatting, enabling zero-shot capabilities like figure-ground segmentation and edge detection while preserving MAE's representation quality.
This paper explores Masked Autoencoders (MAE) with Gaussian Splatting. While reconstructive self-supervised learning frameworks such as MAE learns good semantic abstractions, it is not trained for explicit spatial awareness. Our approach, named Gaussian Masked Autoencoder, or GMAE, aims to learn semantic abstractions and spatial understanding jointly. Like MAE, it reconstructs the image end-to-end in the pixel space, but beyond MAE, it also introduces an intermediate, 3D Gaussian-based representation and renders images via splatting. We show that GMAE can enable various zero-shot learning capabilities of spatial understanding (e.g., figure-ground segmentation, image layering, edge detection, etc.) while preserving the high-level semantics of self-supervised representation quality from MAE. To our knowledge, we are the first to employ Gaussian primitives in an image representation learning framework beyond optimization-based single-scene reconstructions. We believe GMAE will inspire further research in this direction and contribute to developing next-generation techniques for modeling high-fidelity visual data. More details at https://brjathu.github.io/gmae