BlinkFlow
收藏资源简介:
BlinkFlow是由浙江大学计算机辅助设计与图形学国家重点实验室开发的用于事件相机光学流估计的大规模数据集。该数据集通过BlinkSim模拟器生成,包含33000条多样化和挑战性的数据,涵盖了复杂的运动模式和场景类别,并提供了精确的光学流地面实况。BlinkFlow旨在推动事件相机在光学流估计领域的应用,解决现有数据集规模有限和偏差问题,为深度学习方法提供丰富的训练和评估资源。
BlinkFlow is a large-scale dataset for event-based optical flow estimation, developed by the State Key Laboratory of Computer-Aided Design and Computer Graphics at Zhejiang University. It is generated via the BlinkSim simulator, containing 33,000 diverse and challenging data samples that cover complex motion patterns and scene categories, and equipped with accurate optical flow ground truth. BlinkFlow aims to advance the application of event cameras in the field of optical flow estimation, addressing the limitations of limited scale and bias issues in existing datasets, and providing abundant training and evaluation resources for deep learning methods.

- 1BlinkFlow: A Dataset to Push the Limits of Event-based Optical Flow Estimation浙江大学计算机辅助设计与图形学国家重点实验室 · 2023年



