KITTI-FC 和 GoPro-FC
收藏资源简介:
KITTI-FC 和 GoPro-FC 是由浙江大学和湖南大学共同创建的光流估计鲁棒性基准数据集。这些数据集包含200对图像,专门用于评估光流模型在各种常见损坏情况下的鲁棒性。数据集的创建过程包括引入7种时间损坏和17种经典单图像损坏,并通过先进的PSF模糊模拟方法进行增强。这些数据集主要应用于自动驾驶和视频编辑领域,旨在解决光流模型在实际应用中的鲁棒性问题。
KITTI-FC and GoPro-FC are robust benchmark datasets for optical flow estimation jointly created by Zhejiang University and Hunan University. Comprising 200 image pairs, these datasets are specifically tailored to evaluate the robustness of optical flow models under a wide range of common corruption conditions. During their development, 7 temporal corruptions and 17 classic single-image corruptions are introduced, with additional enhancements implemented via advanced PSF blur simulation approaches. These datasets are mainly deployed in the domains of autonomous driving and video editing, with the core objective of addressing the robustness challenges of optical flow models in real-world scenarios.




