Cucumbers, valued for their high water content and nutrient density, offer various health benefits including blood sugar regulation, constipation prevention, and weight loss support, while their appealing taste further contributes to their popularity among consumers. However, farmers worldwide face challenges in accurately identifying various cucumber diseases, which significantly impacts their economic stability, crop yield, and quality. Early detection of these diseases is crucial for maintaining the financial health of agricultural operations. Traditional manual methods for diagnosing cucumber diseases are often time-consuming, subjective, and labor-intensive. A Cucumber Disease Recognition Dataset, which includes 6400 samples across eight different classes, was used for our proposed system. In this study, deep transfer learning models, specifically Convolutional Neural Networks (CNNs), were employed to detect various cucumber diseases. These models were trained and validated by fine-tuning hyperparameters such as the number of layers, loss function, activation function, learning rate, and number of epochs. Among the CNN-based frameworks tested, including MobileNetV2, Xception, InceptionV3, VGG19, and ResNet50, the customized DenseNet169 model, referred to as DenseCucumberNet, demonstrated superior performance. To visualize model classification results, the Grad-CAM technique, an Explainable Artificial Intelligence (XAI) method, was applied to the final layer of the models. The DenseCucumberNet model achieved the highest training accuracy of 98.74% and validation accuracy of 98.16%, with minimal loss.