Machine learning (ML) provides a great opportunity for the construction of models with improved accuracy for classical molecular dynamics (MD). However, the accuracy of a machine learning trained model is limited by the quality and quantity of the training data. Generating large sets of accurate ab-initio training data can require significant computational resources. Furthermore, inconsistent or incompatible data with different accuracies obtained using different methods may lead to biased or unreliable ML models that do not accurately represent the underlying physics. Recently, transfer learning showed its potential for avoiding these problems as well as for improving the accuracy, efficiency, and generalization of ML models using multi-fidelity data. In this work, ab-initio trained ML-based MD (aML-MD) models are developed by transfer learning using DFT data from multiple sources with varying accuracy within the Deep Potential MD (DPMD) framework. The accuracy of the force field is demonstrated by calculating rate constants for the H + HO2 -> H2 + O2(triplet) reaction using quasi-classical trajectories (QCT). We show that the aML-MD model with transfer learning can accurately predict the rate constants while reducing the computational cost by more than 5 times compared to the use of more expensive quantum chemistry training datasets. Hence, the aML-MD model with transfer learning shows great potential in using multi-fidelity data to reduce the computational cost involved in generating the training set for these potentials.