The dataset contains 1623 characters in 50 languages, and each character consists of an aedata4 file and a CSV file.
Each aedata4 file contains 20 samples with the form of (timestamp, x, y, polarity), and CSV file marks the beginning and ending timestamps. Files’ names mean the sequence number of related character.
For more information of how the dataset was created, please refer to the paper.
We provide an interface to this dataset so that users can easily access their own applications using Pytorch, Python 3 is recommended.
Preprocessing code for the dataset can be found here: https://github.com/Brain-Cog-Lab/N-Omniglot.
The program will first read the aedat4 file and split the file contents when first using the code.
More details can be found in this program.
[1] Yang Li, Yiting Dong, Dongcheng Zhao, Yi Zeng. N-Omniglot: a Large-scale Dataset for Spatio-temporal Sparse Few-shot Learning. figshare (2021).
[2] Yang Li, Yiting Dong, Dongcheng Zhao, Yi Zeng. N-Omniglot: a Large-scale Neuromorphic Dataset for Spatio-temporal Sparse Few-shot Learning. Scientific Data, 9(746), Nature Publishing Group, 2022.