N-Omniglot: a Large-scale Neuromorphic Dataset for Spatio-temporal Sparse Few-shot Learning

N-Omniglot is a large neuromorphic few-shot learning dataset. It reconstructs strokes of Omniglot as videos and uses Davis346 to capture the writing of the characters. The recordings can be displayed using DV software's playback function (https://inivation.gitlab.io/dv/dv-docs/docs/getting-started.html). N-Omniglot is sparse and has little similarity between frames. It can be used for event-driven pattern recognition, few-shot learning and stroke generation. It is a neuromorphic event dataset composed of 1623 handwritten characters obtained by the neuromorphic camera Davis346. Each type of character contains handwritten samples of 20 different participants. The file structure and sample can be found in the corresponding PNG file. The dataset has been plugged into the BrainCog framework. <strong>Project</strong>: https://www.brain-cog.network/dataset/N-Omniglot/ <strong>Code</strong>: https://github.com/Brain-Cog-Lab/N-Omniglot <strong>Paper</strong>: https://www.nature.com/articles/s41597-022-01851-z

N-Omniglot是一款大型神经形态（neuromorphic）少样本学习数据集。该数据集将Omniglot数据集的手写笔画重构为视频，并借助Davis346采集字符的书写过程。采集得到的记录数据可通过DV软件的播放功能进行展示（https://inivation.gitlab.io/dv/dv-docs/docs/getting-started.html）。N-Omniglot数据集具有数据稀疏性，且帧间相似度极低，可应用于事件驱动模式识别、少样本学习以及笔画生成等任务。该数据集属于神经形态事件数据集，共包含1623类手写字符，所有数据均由神经形态相机Davis346采集得到。每类字符均包含20名不同参与者的手写样本。数据集的文件结构与样本样例可在配套的PNG文件中查阅。本数据集已集成至BrainCog框架中。<strong>项目主页</strong>：https://www.brain-cog.network/dataset/N-Omniglot/ <strong>代码仓库</strong>：https://github.com/Brain-Cog-Lab/N-Omniglot <strong>相关论文</strong>：https://www.nature.com/articles/s41597-022-01851-z