AI-driven detection of tiny pests in foliage: Integrating image processing and deep learning

We present a novel computer vision method for detecting insect pests on plant and tree leaves under real-world conditions, combining deep learning with classical image processing techniques. Detecting small, sparsely distributed, or camouflaged insects is challenging, as current state-of-the-art object detection methods, primarily designed for larger objects, often overlook them. Our approach to this problem is twofold. First, we employ a deep learning model to analyze suspicious leaves for anomalies (a task well suited to deep learning). However, since deep models struggle with tiny objects in complex backgrounds, we complement them with conventional image processing to pre-identify potentially infested foliage, guiding the model toward relevant areas and mitigating its limitations. This combined strategy proves effective and competitive with other methods across diverse datasets and real-world scenarios. Furthermore, we also conduct a detailed analysis to interpret the model’s predictions, strengthening confidence in its effectiveness.