Artificial Intelligence (AI) and machine learning (ML)-based applications in the field of image processing prove beneficial for the identification and classification of soil types. This study focuses on accurately processing soil images, employing the deep learning (DL) Tensorflow and Keras frameworks in conjunction with cutting-edge deep learning models. Numerous ML models have been previously deployed for soil image classification. The dataset comprises 312 soil images representing five distinct soil types (Black, Cinder, Laterite, Peat, and Yellow), partitioned into training and test sets. Augmentation techniques were applied to enhance the dataset, and subsequent training of models utilized these augmented images. Six Deep Convolution Neural Network (DCNN) models—VGG-16, ResNet150, MobileNet-V2, Inception-ResNetV2, EfficientNet, and DenseNet201—were implemented for soil image classification. The respective accuracies of ResNet150, VGG-16, MobileNet-V2, Inception-ResNetV2, EfficientNet, and DenseNet201 models were 89.28%, 92.85%, 78.57%, 91.07%, 87.50%, and 94.64%. Performance assessment involved precision, F1-score, recall, accuracy, and confusion matrix techniques. Notably, the DenseNet201 model exhibited superior performance as compared to the other state-of-the-art five DCNN models.