Preliminary Result in the Detection of Four Varieties of Mangoes with YOLOv11: Amélie, Kent, Mangot vert, Brooks
- Lecture Notes of the Institute for Computer Sciences, Social Informatics and Telecommunications Engineering (LNICST) , 173 : 178-190
Résumé
The automatic detection of mango varieties using YOLOv11 was explored in order to evaluate the performance of this deep learning model in an agricultural context in Burkina Faso. The study focused on four representative varieties: Brooks(Retard), Mangot vert, Amélie, and Kent. Average Precision (AP) was chosen as the main metric, since it combines both precision and recall, providing a comprehensive evaluation of the model’s detection and classification ability.
The results reveal a high AP for Brooks(Retard) (73.6%) and Mangot vert (71.13%), indicating effective differentiation of these varieties, which display distinctive visual characteristics. In contrast, distinguishing between Amélie (70.9%) and Kent (64.4%) remains more challenging due to their strong morphological and chromatic similarity, which leads to greater confusion during classification.
Overall evaluation indicates a mAP of 70%, reflecting satisfactory performance under real orchard conditions, despite inter-variety variability, changes in lighting, and the diversity of shooting angles. These results highlight the robustness of the model and confirm its potential for practical application in orchard monitoring.
Mots-clés
mango, YOLOv11, object detection, Amélie, Kent, orchard monitoring, fruit classification, Burkina Faso