Leveraging M‐CAR4Net and Hybrid Algorithms for Accurate Tomato Leaf Disease Detection in Complex Environments

Tomatoes are very valuable both economically and nutritionally. However, the prevalence of illnesses may have a significant impact on the quality and output. The quantity, quality, and productivity of the relevant product are all negatively impacted. It is crucial to check the tomato crop leaves for disease symptoms because these issues, if ignored, can damage the plant and result in significant losses in the selling fields. When taking pictures of tomato disease against complex backgrounds, environmental disturbances may make it difficult to get reliable detection and identification results. In addition to powdery mildew, the most prevalent tomato diseases examined in this study's analysis of disease imagery in complex contexts are bacterial spot, mold, gray leaf spot, and healthy. It addresses issues like excessive interference, erroneous lesion localization for small targets, and high rates of false positives and false negatives in actual tomato growing environments. The modified rider optimization algorithm is used to fine‐tune the suggested model's limits, and the modified convolutional attention Res4Net (M‐CAR4Net) model—which is associated with numerous disorders—is used to retrieve the most important features. Finally, the classification is performed using a weighted feature fusion‐based pyramid network (WFFPN). A proposed hybrid version of the gazelle‐Nelder–Mead (GOANM) method is used to refine the proposed classifier. The results indicate that the suggested classical model outperforms all other models on the publicly available leaf disease data samples. Our algorithm performs better than previous and existing studies in detecting diseases affecting tomato plant leaves.