Accurate soil analysis is crucial for optimizing crop cultivation and management since soil quality and texture are so important in the agricultural industry. We have incorporated deep learning (DL) into agriculture for classifying different soil textures, to enable farmers to increase crop output. Convolutional Neural Networks (CNNs), a powerful DL framework, are used in our method to automatically recognize silts, stones, and other color categories by learning detailed patterns inside soil images. The implementation is leveraged in Keras, an advanced neural network interface that is powered by TensorFlow for model creation and training. We harnessed the power of the Image Generator library to efficiently manage the image dataset. We experimented with the evaluation of four CNN models: Baseline, Augment, Dropout, and Padding. The Baseline model achieved an accuracy of 82%. The Augment model, which incorporated image augmentation, delivered an accuracy of 78%, while the Dropout model, utilizing dropout layers, achieved 60%. Notably, the Padding model excelled, yielding an impressive accuracy of 90%. This study marks a significant stride in harnessing DL technology for the betterment of agriculture, offering farmers the tools they need to make informed decisions and enhance their crop yields sustainably.