Mango leaf diseases critically impact both quality and yield in global mango cultivation. Traditional detection methods, reliant on manual inspection, are labor-intensive and error-prone. The proposed model successfully mitigates the gradient vanishing problem, enhancing training stability and improving overall model performance. This work introduces a novel method employing a deep residual skip-connected convolutional neural network (DRCNN) for the identification and categorization of diseases affecting mango leaves. Extensive analysis of the DRCNN, including measures like batch sizes and learning rates, revealed that at 0.001 learning rate and batch size of 8, the network yielded the highest performance. This configuration achieved a remarkable precision of 99.76%, an accuracy of 99.75%, an F1-score of 0.9975, and a recall of 0.9976. Results on the mango leaf dataset demonstrate the proposed architecture’s superior accuracy over traditional and contemporary techniques. By automating disease detection, this framework facilitates timely and precise management, significantly enhancing mango production's productivity and sustainability.