PakSoil DataSet : Controlled 55 MP Micro-Images, Scale-Reference Images, and Mobile Images for Soil Type Classification

PakSoil @ PakHPCThis record contains a curated public subset of the soil-image dataset associated with the manuscript “VMSA: Soil type identification using a visual micro-feature-based soil analyzer.” The dataset supports image-based soil classification research using controlled high-resolution imaging and hand-crafted visual descriptors. The record includes images from four soil classes: - Clay - Loam - Sand - Sandy Loam The files are organized into three acquisition groups: 0_Micro_55MP: controlled high-resolution micro-images acquired using a 55 MP imaging setup. These images are the closestpublic subset to the paper’s main visual benchmark setting.1_scale_Micro_55MP: scale-reference images acquired under the same controlled setup to support calibration andinterpretation of particle-scale measurements.2_Mobile: supplementary mobile-phone images included for qualitative comparison and future field-oriented extensions. This upload is intended to improve transparency and reproducibility of the VMSA framework. The public subset can be used for qualitative inspection, method understanding, and partial replication of the reported workflow. The paper combines particle morphology, texture, and CIE Lab colour descriptors with a one-dimensional convolutional neural network (HCFA-CNN) for soil type classification under controlled acquisition conditions. The record includes compressed image folders together with metadata and documentation files such as a README, manifest, class summary, and citation information.