Image Classification2009en
Canadian Institute for Advanced Research 100
60K 32x32 color images in 100 fine-grained classes grouped into 20 superclasses. More challenging than CIFAR-10.
Current State of the Art
ViT-H/14
94.55
accuracy
accuracy Progress Over Time
Showing 3 breakthroughs from Aug 2017 to Oct 2020
Key Milestones
Total Improvement
21.2%
Time Span
3y 3m
Breakthroughs
3
Current SOTA
94.5
Top Models Performance Comparison
Top 4 models ranked by accuracy
Best Score
94.5
Top Model
ViT-H/14
Models Compared
4
Score Range
16.5
accuracyPrimary
| # | Model | Score | Paper / Code | Date |
|---|---|---|---|---|
| 1 | ViT-H/14Open Source Google | 94.55 | Dec 2025 | |
| 2 | EfficientNet-B7Open Source Google | 91.7 | Dec 2025 | |
| 3 | ViT-B/16Open Source Google | 91.48 | Dec 2025 | |
| 4 | ResNet-50Open Source Microsoft | 78.04 | Dec 2025 |