Image Classification On Kuzushiji Mnist

Métriques

Accuracy

Error

Résultats

Résultats de performance de divers modèles sur ce benchmark

Nom du modèle	Accuracy	Error	Paper Title	Repository
VGG8B(2x) + LocalLearning + CO	99.01	0.99	Training Neural Networks with Local Error Signals
ResNet-14	98.75	-	CNN Filter DB: An Empirical Investigation of Trained Convolutional Filters
NSRL (log D) (d=32)	98.63	-	Toward Understanding Supervised Representation Learning with RKHS and GAN	-
NSRL (log D) (d=8)	98.61	-	Toward Understanding Supervised Representation Learning with RKHS and GAN	-
KMNIST-Tiny	99.35	-	Efficient Global Neural Architecture Search
linear/flexible model	79.90	-	Multi-Complementary and Unlabeled Learning for Arbitrary Losses and Models	-
Convolutional Tsetlin Machine	96.3	-	The Convolutional Tsetlin Machine
Resnet-152	98.79	-	A Comprehensive Study of ImageNet Pre-Training for Historical Document Image Analysis	-
KMNIST-Mobile	99.29	-	Efficient Global Neural Architecture Search
PreActResNet-18 + Input Mixup	98.41	-	mixup: Beyond Empirical Risk Minimization
FWD	79.5	-	Multi-Complementary and Unlabeled Learning for Arbitrary Losses and Models	-
ResNet18 + VGG Ensemble	-	1.10	Deep Learning for Classical Japanese Literature
CN(d=8)	98.60	-	Toward Understanding Supervised Representation Learning with RKHS and GAN	-
Efficient Capsnet	98.43	-	Improved efficient capsule network for Kuzushiji-MNIST benchmark dataset classification
Complementary-Label Learning	67.1	-	Complementary-Label Learning for Arbitrary Losses and Models
NSRL (WGAN) (d=8)	98.68	-	Toward Understanding Supervised Representation Learning with RKHS and GAN	-
KerCNN	93.13	-	KerCNNs: biologically inspired lateral connections for classification of corrupted images	-
CAMNet3	99.05	0.95	Context-Aware Multipath Networks	-
CN(d=16)	98.80	-	Toward Understanding Supervised Representation Learning with RKHS and GAN	-
NSRL (log D) (d=16)	98.81	-	Toward Understanding Supervised Representation Learning with RKHS and GAN	-

0 of 26 row(s) selected.