HiVision: Rapid visualization of large-scale spatial vector data

Rapid visualization of large-scale spatial vector data is a long-standing challenge in Geographic Information Science. In existing methods, the computation overheads grow rapidly with volumes, leading to incapability providing real-time for data, even parallel acceleration technologies. To fill gap, we present HiVision, display-driven model data. Different from traditional data-driven computing units HiVision are pixels rather than objects achieve performance, and efficient spatial-index-based strategies introduced estimate topological relationships between objects. can maintain exceedingly good performance regardless volume due stable pixel number display. addition, an optimized architecture proposed ensure ability visualization. Experiments show that our approach outperforms methods rendering speed visual effects while dealing provide interactive datasets billion-scale points/segments/edges flexible styles. The code open-sourced at https://github.com/MemoryMmy/HiVision online demonstration.