<p>With the advancement of mining depth, microseismic monitoring technology has become increasingly important in mine safety management, earthquake early warning, and rock engineering stability assessment. However, in practical applications, microseismic data are usually accompanied by strong noise and complex waveform characteristics, making traditional pickup methods face greater challenges in terms of accuracy and robustness. Especially in the mining environment, the influence of noise makes the quality and scale of microseismic data poor, further increasing the difficulty of P-wave arrival pickup. Aiming at the problems of the traditional microseismic monitoring model, which is large in volume, complex in computation, and slow in processing speed, this paper proposes a microseismic P-wave arrival pickup method that integrates lightweight double adaption mixed scale-net (DAMS-Net) and transfer learning. The computational volume and model complexity are reduced by the DAMS-Net, combined with transfer learning to enhance the robustness of the model in the presence of large noise and data imbalance. The soft-label labeling method with S-Laplace distribution is introduced further to improve the accuracy and generalization ability of the model. Experimental results show that the proposed method significantly outperforms the traditional method in terms of accuracy and robustness, and the usage of memory resources has decreased by more than 40%. The method not only improves the accuracy of P-wave arrival pickup but also effectively reduces the dependence on a large amount of labeled data, which has strong practical application value and provides an efficient and accurate solution for mine microseismic monitoring and mine safety early warning.</p>