Soil Texture Classification Using RESNET-152

Soil texture classification is vital for agriculture and soil management as it affects crop productivity and environmental sustainability. Conventional classification techniques usually involve manual assessment, which is tedious work that requires specialized skills. This study concerns a CNN- based deep learning approach with ResNet-152 architecture aimed at classifying soil pictures into sand, silt, or clay. The model was trained with a dataset of soil pictures to automatically extract texture features and analyze them while achieving high accuracy of classification rate. In contrast to traditional approaches, the proposed system, besides accomplishing soil texture classification, actually predicts the exact percentage composition of sand, silt, and clay in a sample envisaged. The further portion of the system is a crop recommendation module that will recommend suitable crops based on the soil texture composition, thus maximizing the agricultural decision-making process. Average classification accuracy of the model stands at 90 percent.