Curvature-Aware Training for Coordinate Networks
This addresses a bottleneck for real-time applications in computer vision by improving training efficiency without sacrificing memory, though it is incremental as it builds on existing optimization techniques.
The paper tackles the slow training of coordinate networks with first-order optimizers by using second-order optimization methods, reducing training times while maintaining compressibility, as demonstrated across audio, images, videos, shape reconstruction, and neural radiance fields.
Coordinate networks are widely used in computer vision due to their ability to represent signals as compressed, continuous entities. However, training these networks with first-order optimizers can be slow, hindering their use in real-time applications. Recent works have opted for shallow voxel-based representations to achieve faster training, but this sacrifices memory efficiency. This work proposes a solution that leverages second-order optimization methods to significantly reduce training times for coordinate networks while maintaining their compressibility. Experiments demonstrate the effectiveness of this approach on various signal modalities, such as audio, images, videos, shape reconstruction, and neural radiance fields.