Agriculture is very important in Africa. But farmers are dealing with many plants pathologies that keep agriculture from developing and boosting the economy. Current pathologies diagnosis based on human screening is time consuming and expensive, while computer visionbased models are efficiently promising. That is why we address the problem of identification of the pathologies affecting agricultural crops with computer vision tools. Though the high variability of symptoms due to the age of infected tissues, genetic differences, and light conditions in trees reduces the precision of detection, we propose a method to achieve our goal.