osiriszjq / complex_encoding

Trading Positional Complexity vs Deepness in Coordinate Networks

Trading Positional Complexity vs Deepness in Coordinate Networks

Project Page | Paper

Jianqiao Zheng, Sameera Ramasinghe, Xueqian Li, Simon Lucey
The University of Adelaide

This is the official implementation of the paper "Trading Positional Complexity vs Deepness in Coordinate Networks", which has been accepted to ECCV 2022.

Illustration of different methods to extend 1D encoding

Google Colab

If you want to try out our new complex encoding, we have written a Colab with the following experiments:

• simple encoding for 2D image reconstuction with separable coordinates
• complex encoding for 2D image reconstuction with separable coordinates
• colsed form solution of complex encoding for 2D image reconstuction with separable coordinates.
• simple encoding for 3D video reconstuction with non-separable coordinates
• complex encoding for 3D video reconstuction with non-separable coordinates

Dataset

The Dataset used to reproduced can be found in Google Drive. The image data is from Random Fourier Frequency and the video data is from Youtube.

Citation

@inproceedings{zheng2022trading,
    title={Trading Positional Complexity vs Deepness in Coordinate Networks},
    author={Zheng, Jianqiao and Ramasinghe, Sameera and Li, Xueqian and Lucey, Simon},
    booktitle={Proceedings of the European Conference on Computer Vision (ECCV)},
    year={2022}
}