60,000 color images of 32x32 pixels categorized into 100 distinct classes and 20 superclasses. The dataset is partitioned into 50,000 training images and 10,000 test images, with each class containing exactly 600 instances.
Use Cases
- Train image classification models using the 'fine_labels' to distinguish between specific categories like 'beaver' and 'otter'
- Implement hierarchical learning algorithms by mapping 'fine_labels' to their corresponding 'coarse_labels'
- Benchmark small-scale convolutional neural networks (CNNs) on 3072-byte 'data' arrays representing flattened RGB pixels
Strengths
- 60,000 32x32 RGB images across 100 fine-grained categories
- Hierarchical structure featuring 20 superclasses (coarse labels) such as 'aquatic mammals' and 'trees'
- Balanced distribution with exactly 500 training and 100 testing images per class
- Python-specific pickle format containing 'data', 'fine_labels', and 'coarse_labels' keys