Table 5 Specifications and descriptions of image classification datasets used in this study
From: Extreme learning machine versus classical feedforward network
Dataset | Size | Images examples | Description |
|---|---|---|---|
MNIST OCR | 70,000 |
| Recognition of handwritten digits. Image resolution: 28\(\times\)28 px, Black and white images, Number of classes: 10 |
CIFAR-10 | 60,000 |
| Recognition of 10 various classes of objects on photographs. Image resolution: 32\(\times\)32 px, RGB colour images, Number of classes: 10 |
Small NORB | 48,600 |
| Classification of objects photographed from various angles and with varying lighting. Image resolution: 96\(\times\)96 px, Grayscale images, Number of classes: 5 |
GTSRB | 51,839 |
| Recognition of German road signs. Image resolution: various, we rescaled them to the median resolution: 43\(\times\)43 px, RGB colour images, Number of classes: 43 |
ImageNet | 1.2 mln |
| Classification of various objects. Image resolution: 224\(\times\)224, rescaled to two variants: 16\(\times\)16 px and 32\(\times\)32 px, RGB colour images, Number of classes: 1000 |
