Datasets › Kuzushiji-MNIST

Kuzushiji-MNIST

Introduced by Tarin Clanuwat et al. in Deep Learning for Classical Japanese Literature archive 2025-07-28

Kuzushiji-MNIST is a drop-in replacement for the MNIST dataset (28x28 grayscale, 70,000 images). Since MNIST restricts us to 10 classes, the authors chose one character to represent each of the 10 rows of Hiragana when creating Kuzushiji-MNIST. Kuzushiji is a Japanese cursive writing style.

Source: Deep Learning for Classical Japanese Literature Image Source: https://github.com/rois-codh/kmnist

Homepage Benchmarks Papers JSON

Benchmarks archive 2025-07-28

All 2 leaderboards whose dataset resolves to this page shown (sort by any header). "First row" is the archive's own first row at snapshot, in the archive's row order; nothing here re-ranks and metric direction is not asserted.

		First row (archive order)	Paper	Code
Image Classification	Kuzushiji-MNIST	KMNIST-Tiny Accuracy 99.35	Efficient Global Neural Architecture Search	siddikui/Efficient-Macro-Micro-NAS	26	Compare
Fine-Grained Image Classification	Kuzushiji-MNIST	VGG-5 Accuracy 98.98	ProgressiveSpinalNet architecture for FC layers	praveenchopra/ProgressiveSpinalNet	1	Compare

Papers archive 2025-07-28

17 shown of 17 papers with a leaderboard row on this dataset's benchmarks, newest first. The archive's own "papers using this dataset" list was never published, so this is the benchmark-backed subset; the archive's count for this dataset is 97. The Syntology column is from Syntology's graph (read 2026-09-24), stated per sample; it is not part of any archive number.

			Date	Samples run Syntology
Efficient Global Neural Architecture Search	1	2	8 Feb 2025	not harvested
Learning local discrete features in explainable-by-design convolutional neural networks	1	1	31 Oct 2024	not harvested
Improved efficient capsule network for Kuzushiji-MNIST benchmark dataset classification	1	1	15 Dec 2023	not harvested
CNN Filter DB: An Empirical Investigation of Trained Convolutional Filters	1	1	29 Mar 2022	not harvested
ProgressiveSpinalNet architecture for FC layers	1	1	21 Mar 2021	not harvested
Toward Understanding Supervised Representation Learning with RKHS and GAN	0	9	1 Jan 2021	not harvested
SpinalNet: Deep Neural Network with Gradual Input	3	1	7 Jul 2020	not harvested
Multi-Complementary and Unlabeled Learning for Arbitrary Losses and Models	0	2	13 Jan 2020	not harvested
KerCNNs: biologically inspired lateral connections for classification of corrupted images	0	1	18 Oct 2019	not harvested
Context-Aware Multipath Networks	0	1	26 Jul 2019	not harvested
The Convolutional Tsetlin Machine	9	1	23 May 2019	ran 0 of 1 samples (1 unverified)
A Comprehensive Study of ImageNet Pre-Training for Historical Document Image Analysis	0	1	22 May 2019	not harvested
Training Neural Networks with Local Error Signals	2	1	20 Jan 2019	not harvested
Deep Learning for Classical Japanese Literature	10	1	3 Dec 2018	ran 2 of 11 samples (9 unverified; 1 pointer-only for licence)
Complementary-Label Learning for Arbitrary Losses and Models	1	1	10 Oct 2018	ran 0 of 6 samples (6 unverified)
mixup: Beyond Empirical Risk Minimization	71	1	25 Oct 2017	ran 30 of 47 samples (17 unverified; 15 pointer-only for licence)
Identity Mappings in Deep Residual Networks	54	1	16 Mar 2016	ran 3 of 25 samples (22 unverified; 1 pointer-only for licence)

Dataset loaders archive 2025-07-28

docs.activeloop.ai/datasets/kmnist activeloopai/Hubtfpytorch

www.tensorflow.org/datasets/catalog/kmnist tensorflow/datasetstfjax

rois-codh/kmnist

3 loaders as listed in the archive; links are outbound and not re-checked here.

Tasks archive 2025-07-28

Image Classification Fine-Grained Image Classification

License archive 2025-07-28

CC BY-SA 4.0

Modalities archive 2025-07-28

Images

Languages archive 2025-07-28

Japanese

Variants archive 2025-07-28

Kuzushiji-MNIST

1 variant name, as the archive lists them.

Report a problem or propose a change · a person checks every report against the paper or source before anything changes; decisions are listed on /corrections