EventSTR
收藏资源简介:
EventSTR是一个基于事件相机构建的大规模高清晰度场景文本识别数据集,由安徽大学计算机科学与技术学院的研究者们收集和注释。该数据集包含9928个样本,覆盖了中英文字符,能够在不同的照明条件、运动、遮挡等复杂环境中有效支持研究。数据集的构建旨在推动事件相机在各种行业中的应用,特别是在场景文本识别领域,以解决传统RGB相机在低光照、复杂背景、运动模糊等挑战因素下的性能限制。
EventSTR is a large-scale high-definition scene text recognition dataset built based on event cameras, collected and annotated by researchers from the School of Computer Science and Technology, Anhui University. This dataset contains 9,928 samples covering both Chinese and English characters, and can effectively support research in complex environments such as varying lighting conditions, motion scenarios and occlusions. The dataset is constructed to promote the application of event cameras in various industries, especially in the field of scene text recognition, so as to address the performance limitations of traditional RGB cameras under challenging factors such as low-light conditions, complex backgrounds and motion blur.

- 1EventSTR: A Benchmark Dataset and Baselines for Event Stream based Scene Text Recognition安徽大学计算机科学与技术学院 · 2025年



