Implicit Neural Representation with Multi-Scale Sine Activation

Implicit Neural Representation (INR) has become a powerful paradigm for modeling continuous signals in computer vision, graphics, and scientific computing. However, neural networks generally suffer from severe spectral bias, which limits their ability to accurately model high-frequency details and multi-scale structures. To address this challenge, we propose a novel Multi-Scale Sine Activation (MSA) function, which explicitly introduces multi-scale frequency responses by incorporating multiple sets of sine activations with logarithmically spaced frequencies in parallel at each layer. MSA is further combined with an amplitude modulation mechanism to ensure numerical stability and robust optimization across different frequency channels. We conduct extensive experiments on a series of challenging tasks, including 1D high-frequency and multi-scale function fitting, 2D image fitting, video fitting, 3D shape representation, and PDEs solving. Experimental results show that MSA-Net outperforms existing mainstream INR methods in terms of reconstruction accuracy, detail preservation, and training stability.