This paper presents an improved event-based stereo visual odometry (ESVO) system that enhances accuracy and efficiency. The proposed method addresses the limitations of existing approaches by introducing an adaptive accumulation (AA) strategy for event data, which efficiently determines pixel locations associated with instantaneous edges. It also improves the mapping sub-problem by combining temporal and static stereo configurations, and introduces an IMU-aided solution for camera pose tracking to overcome the insensitivity to the yaw component of general 6-DoF motion. The system is evaluated on publicly available datasets, demonstrating better performance in terms of mapping completeness, depth smoothness, and trajectory accuracy. The proposed method achieves more accurate and complete reconstruction results, especially for horizontal edges, and significantly improves computational efficiency by reducing the number of points used for mapping. The system is implemented as an open-source software for future research in this field. The results show that the proposed method outperforms the original ESVO in both quantitative and qualitative evaluations, with improved accuracy and efficiency in trajectory estimation and mapping. The system is also shown to be more robust in handling dynamic scenarios, particularly when inertial measurements are used as a prior. The paper concludes that the proposed method represents a significant advancement in event-based visual odometry, offering a more efficient and accurate solution for practical applications.This paper presents an improved event-based stereo visual odometry (ESVO) system that enhances accuracy and efficiency. The proposed method addresses the limitations of existing approaches by introducing an adaptive accumulation (AA) strategy for event data, which efficiently determines pixel locations associated with instantaneous edges. It also improves the mapping sub-problem by combining temporal and static stereo configurations, and introduces an IMU-aided solution for camera pose tracking to overcome the insensitivity to the yaw component of general 6-DoF motion. The system is evaluated on publicly available datasets, demonstrating better performance in terms of mapping completeness, depth smoothness, and trajectory accuracy. The proposed method achieves more accurate and complete reconstruction results, especially for horizontal edges, and significantly improves computational efficiency by reducing the number of points used for mapping. The system is implemented as an open-source software for future research in this field. The results show that the proposed method outperforms the original ESVO in both quantitative and qualitative evaluations, with improved accuracy and efficiency in trajectory estimation and mapping. The system is also shown to be more robust in handling dynamic scenarios, particularly when inertial measurements are used as a prior. The paper concludes that the proposed method represents a significant advancement in event-based visual odometry, offering a more efficient and accurate solution for practical applications.