This article presents a carefully curated dataset of mango leaves, collected from the northern regions of Bangladesh, specifically from Naogaon District, a major mango-producing area. Bangladesh, known for its agriculture, frequently faces leaf diseases that impact mango yield and quality. Early detection is crucial to prevent widespread damage and ensure better disease management. The dataset comprises 4921 raw image samples, categorized into five distinct classes: Healthy, Anthracnose, Powdery Mildew, Turning Brown, and Gall Midge. The images were captured under natural lighting conditions to preserve the leaves' intrinsic visual features, ensuring authenticity and variability. This dataset is a valuable resource for botanical research and machine learning applications, particularly in the automated classification of mango leaf diseases, helping researchers develop more effective disease detection models.