arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2307.05033cs.CVcs.ROeess.IV
收藏

利用事件相机实现随时光流估计

Towards Anytime Optical Flow Estimation with Event Cameras

Zhejiang University(浙江大学);Hunan University(湖南大学)

Yaozu Ye, Hao Shi, Kailun Yang, Ze Wang, Xiaoting Yin, Lei Sun, Yaonan Wang, Kaiwei Wang

AI总结:

针对事件光流真值帧率低的问题,提出 Unified Voxel Grid 表示、EVA-Flow 随时估计网络和 RFWL 无监督损失,实现 5ms 延迟、200Hz 稠密光流与强泛化。

摘要:

事件相机以毫秒级响应对数亮度变化,因此非常适合光流估计。然而,现有事件相机数据集仅提供低帧率光流真值,限制了事件驱动光流的研究潜力。为应对这一挑战,我们提出一种低延迟事件表示 Unified Voxel Grid(统一体素网格),并提出 EVA-Flow,这是一个基于事件的随时光流估计网络,仅用低帧率光流真值监督即可生成高帧率事件光流。此外,我们提出 Rectified Flow Warp Loss(RFWL,校正光流扭曲损失),用于对中间光流进行无监督评估。在 MVSEC、DESC 以及我们的 EVA-FlowSet 上进行的大量实验表明,EVA-Flow 取得了具有竞争力的性能、超低延迟(5ms)、时间稠密的运动估计(200Hz)以及强泛化能力。代码将在 https://github.com/Yaozhuwa/EVA-Flow 提供。

英文摘要:

Event cameras respond to changes in log-brightness at the millisecond level, making them ideal for optical flow estimation. However, existing datasets from event cameras provide only low frame rate ground truth for optical flow, limiting the research potential of event-driven optical flow. To address this challenge, we introduce a low-latency event representation, Unified Voxel Grid, and propose EVA-Flow, an EVent-based Anytime Flow estimation network to produce high-frame-rate event optical flow with only low-frame-rate optical flow ground truth for supervision. Furthermore, we propose the Rectified Flow Warp Loss (RFWL) for the unsupervised assessment of intermediate optical flow. A comprehensive variety of experiments on MVSEC, DESC, and our EVA-FlowSet demonstrates that EVA-Flow achieves competitive performance, super-low-latency (5ms), time-dense motion estimation (200Hz), and strong generalization. Our code will be available at https://github.com/Yaozhuwa/EVA-Flow.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑