This paper presents a dataset of 4507 annotated images of African plums collected across diverse regions in Cameroon, marking the first dataset specifically designed for AI-driven quality assessment of this fruit. The dataset is categorized into six quality grades: unaffected, bruised, cracked, rotten, spotted, and unripe. These categories represent varying degrees of plum quality, from optimal condition to various defects and ripeness levels. Captured under natural lighting using a consistent smartphone setup, the images were meticulously labeled by agricultural experts, ensuring high annotation accuracy. This resource is valuable for developing and testing computer vision, deep learning-based recognition systems and object detection models in agriculture, enabling automated evaluation of plum quality for commercialization. By offering a comprehensive, culturally relevant dataset focused on a traditionally underrepresented crop, this work supports advancements in precision agriculture, particularly in developing regions. Potential applications include AI-based tools for real-time sorting, defect detection, and quality assurance in the supply chain.