Time delay estimation remains an active research area with broad applications across multiple fields. While conventional Generalized Cross-Correlation (GCC) based approaches focus primarily on errors following a normal distribution, they struggle with large estimation errors that deviate from that distribution due to significant signal shifts.

Coarse2Fine-ResNet is a robust deep learning framework that effectively handles both types of errors by leveraging 2D GCC patterns and a ResNet architecture. Evaluated on multiple datasets including acoustic drone flying, speaker localization, and optical fiber sensing, our method outperforms existing GCC-based approaches and state-of-the-art deep learning models in both estimation accuracy and reduction of large errors.