What it is
Classic image classification benchmark with 60k 32×32 RGB images across 10 object classes, preserving the original CIFAR-10 split. 60,000 32×32 RGB images split into 50k train and 10k test across 10 object categories for small-image classification benchmarking.
Key results
- 60,000 RGB images total
- 50,000 train + 10,000 test
- 10 classes at 32×32 resolution
Useful for benchmarking image classification models, including compact CNNs and TinyML setups.