Abstract

Implicit Neural Representation (INR), which utilizes a neural network to map coordinate inputs to corresponding attributes, is causing a revolution in the field of signal processing. However, current INR techniques suffer from a spectral bias and capacity-convergence gap, resulting in imperfect performance when representing complex signals with multiple frequencies. We have identified that both of these two problems could be addressed by increasing the utilization of definition domain in current activation functions, for which we propose the FINER++ framework by extending existing periodic/non-periodic activation functions to variable-periodic ones. By initializing the bias of the neural network with different ranges, sub-functions with various frequencies in the variable-periodic function are selected for activation. Consequently, the supported frequency set can be flexibly tuned, leading to improved performance in signal representation. We demonstrate the generalization and capabilities of FINER++ with different activation function backbones (Sine, Gauss. and Wavelet) and various tasks (2D image fitting, 3D signed distance field representation, 5D neural radiance fields optimization and streamable INRs), and we show that it improves existing INRs.

Flexible spectral-bias tuning

Experiments

Citation

@article{zhu2024finerplusplus,
    title={FINER++: Building a Family of Variable-periodic Functions for Activating Implicit Neural Representation},
    author={Zhu, Hao and Liu, Zhen and Zhang, Qi and Fu, Jingde and Deng, Weibing and Ma, Zhan and Guo, Yanwen and Cao, Xun},
    url={https://arxiv.org/abs/2407.19434}, 
    year={2024},
}
@inproceedings{liu2024finer,
    title={FINER: Flexible spectral-bias tuning in Implicit NEural Representation by Variable-periodic Activation Functions},
    author={Liu, Zhen and Zhu, Hao and Zhang, Qi and Fu, Jingde and Deng, Weibing and Ma, Zhan and Guo, Yanwen and Cao, Xun},
    booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
    pags={2713--2722},
    year={2024}
}

Acknowledgements

The website template was borrowed from Michaël Gharbi.