This paper presents a large-scale, meticulously validated dataset designed for the classification of fruit quality across three critical categories: fresh, rotten, and formalin-mixed. This dataset provides a comprehensive collection of five fruit types—apple, banana, mango, orange, and grapes—classified into three categories. Ensuring fruit safety is essential, especially in detecting formalin-mixed samples. Formalin is a common preservative used to extend shelf life, but it poses serious health risks. The dataset was collected from April 18, 2024, to July 31, 2024, using three different mobile cameras to introduce diversity in image quality and conditions. Initially, the dataset comprised 10,154 images. To enhance its robustness and applicability, eight data augmentation techniques—rotation (i.e., 45, 60, 90 degree), horizontal flipping, zooming, brightness adjustment, shearing, and Gaussian noise addition—were applied, increasing the total number of images to 81,232. All images were uniformly resized to 512 × 512 pixels to ensure consistency and facilitate integration with machine learning and computer vision models. This dataset supports a variety of applications, including automated fruit classification, food quality control, and real-time monitoring through IoT technology. It offers valuable resources for research and development in automated detection systems, improving food safety and processing efficiency, and addressing the health risks associated with formalin exposure.