OpenSTL
收藏资源简介:
OpenSTL数据集是一个全面的时空预测学习基准,由浙江大学和西湖大学联合开发。该数据集包含14种代表性算法和24种模型,涵盖了从合成移动物体轨迹预测到实际人类动作、驾驶场景、交通流量和天气预报等多个领域。数据集支持的任务范围广泛,从微观到宏观尺度,从合成到真实世界数据。OpenSTL数据集通过提供一个模块化和可扩展的框架,实现了对各种最先进方法的标准化评估,旨在推动时空预测学习领域的发展。
The OpenSTL Dataset is a comprehensive spatio-temporal prediction learning benchmark jointly developed by Zhejiang University and Westlake University. It comprises 14 representative algorithms and 24 models, covering multiple domains ranging from synthetic moving object trajectory prediction to real-world human motion, driving scenarios, traffic flow and weather forecasting. The dataset supports a wide range of tasks, spanning from micro to macro scales, and from synthetic to real-world data. By providing a modular and scalable framework, the OpenSTL Dataset enables standardized evaluation of various state-of-the-art methods, aiming to promote the development of the spatio-temporal prediction learning field.

- 1OpenSTL: A Comprehensive Benchmark of Spatio-Temporal Predictive Learning浙江大学 · 2023年



