SoilClass: A Novel Image Dataset for Soil Classification

The dataset contains a total of 552 images captured using various mobile cameras. The original RGB images have different dimensions due to the usage of various devices. We set the image size to 224x224 for the CNN's training. We divided the dataset into five different classes, such as Type 1, Type 2, Type 3, Type 4, and Type 5. Then each class was divided into three parts: test data, train data, and validation data for model evaluation, where train data, test data, and validation data contain 70%, 20%, and 10% of the whole data, respectively. Soil Type #images Type 1 109 Type 2 96 Type 3 137 Type 4 107 Type 5 103