Image Classification2009en

Canadian Institute for Advanced Research 10

60K 32x32 color images in 10 classes. Classic small-scale image classification benchmark with 50K training and 10K test images.

Samples:60,000
Metrics:accuracy
Paper / WebsiteDownload
Current State of the Art

DeiT-B Distilled

Meta

99.1

accuracy

accuracy Progress Over Time

Showing 2 breakthroughs from Aug 2017 to Dec 2020

95.796.697.698.599.4Aug 2017Dec 2020accuracyDate

Key Milestones

Aug 2017
ResNet-50

With Cutout augmentation.

96.0
Dec 2020
DeiT-B DistilledCurrent SOTA

Near-SOTA on CIFAR-10 with transfer learning.

99.1
+3.2%
Total Improvement
3.2%
Time Span
3y 5m
Breakthroughs
2
Current SOTA
99.1

Top Models Performance Comparison

Top 3 models ranked by accuracy

accuracy1DeiT-B Distilled99.1100.0%2ConvNeXt V2 Base98.799.6%3ResNet-5096.096.9%0%25%50%75%100%% of best
Best Score
99.1
Top Model
DeiT-B Distilled
Models Compared
3
Score Range
3.1

accuracyPrimary

#ModelScorePaper / CodeDate
1
DeiT-B DistilledOpen Source
Meta
99.1Dec 2025
2
ConvNeXt V2 BaseOpen Source
Meta
98.7Dec 2025
3
ResNet-50Open Source
Microsoft
96.01Dec 2025

Other Image Classification Datasets

CIFAR-10 Benchmark - Image Classification | CodeSOTA