BlinkVision
收藏资源简介:
BlinkVision是由浙江大学和Avolution AI合作创建的一个大规模、多样化的基准数据集,旨在填补事件数据和图像在对应任务(如光流、场景流和点跟踪)中的综合基准的空白。数据集包含丰富的模态,包括事件数据和RGB图像,并提供密集的每像素注释,涵盖410个日常类别。数据集通过高质量的资产和渲染工具生成,确保了数据的逼真性,并覆盖了多种自然因素。BlinkVision的应用领域广泛,旨在推动图像和事件相机在对应任务中的算法研究,提供新的观察、实践和见解。
BlinkVision is a large-scale, diverse benchmark dataset co-developed by Zhejiang University and Avolution AI, which aims to fill the gap in comprehensive benchmarks for event data and images across key tasks such as optical flow, scene flow, and point tracking. The dataset offers rich multimodal content, including event data and RGB images, alongside dense per-pixel annotations spanning 410 everyday categories. Generated using high-quality assets and rendering pipelines, BlinkVision ensures high data realism and encompasses a diverse set of natural environmental factors. With broad application prospects, this dataset is designed to promote algorithmic research on image and event cameras for the aforementioned tasks, and provide novel research perspectives, practical implementations and valuable insights.



