KMNIST

许多机器学习研究都集中在生成在基准任务上表现良好的模型，从而提高我们对与这些任务相关的挑战的理解。从 ML 研究人员的角度来看，任务本身的内容在很大程度上是无关紧要的，因此越来越多的人呼吁基准任务更加关注与社会或文化相关的问题。在这项工作中，我们介绍了 Kuzushiji-MNIST，一个专注于 Kuzushiji（草书日语）的数据集，以及两个更大、更具挑战性的数据集 Kuzushiji-49 和 Kuzushiji-Kanji。通过这些数据集，我们希望让机器学习社区进入日本古典文学的世界。

Many machine learning studies focus on developing models that achieve strong performance on benchmark tasks, thereby advancing our understanding of the challenges associated with these tasks. From the perspective of ML researchers, the specific content of the tasks themselves is largely irrelevant, leading to growing calls for benchmark tasks to pay more attention to socially or culturally relevant issues. In this work, we introduce Kuzushiji-MNIST, a dataset focused on Kuzushiji (cursive Japanese), along with two larger, more challenging datasets: Kuzushiji-49 and Kuzushiji-Kanji. Through these datasets, we aim to bring the machine learning community into the world of classical Japanese literature.

搜集汇总

背景与挑战

背景概述

KMNIST是一个专注于识别日本古典草书字符的数据集，旨在为机器学习研究提供更具社会文化相关性的基准任务。该数据集由ROIS-DS Center for Open Data in the Humanities于2018年公开发布。