Modified Xception-RF: A biotic stress prediction model in plant leaves using feature extraction and classification

Agriculture stands as one of the foremost cornerstones of a nation's economy, exerting a direct influence on both food security and international trade. The efficient management and control of plant diseases depend on their early diagnosis, which protects agricultural production and guarantees food security. Incorporating modern technology, such as artificial intelligence (AI) in disease recognition, can revolutionise the agricultural industry, ensuring farmers and consumers a more sustainable and prosperous future. Pre-trained models are used in the study for Feature Extraction (FE), while Random Forest (RF) is used to classify plant illnesses. This paper discusses the significance of plant disease recognition in agriculture and introduces a modified Xception-RF model that replaces the Rectified linear unit (ReLU) activation function with a square root-based Leaky ReLU activation function (SR-LReLU) that could detect disease effectively in terms of performance. The modified Xception model has also reduced the number of parameters in the middle flow by replacing the 8 repetitions with 5 repetitions that could detect disease effectively in terms of reduced time consumption. The outcomes illustrate the possibility of merging Machine Learning (ML) and Deep Learning (DL) approaches for improved plant disease recognition, outperforming other models by obtaining above 99% accuracy with only 8.7 seconds of computation time. The results of this research hold the potential to advance the development of more efficient solutions for early disease detection and control in plants, ultimately benefiting both farmers and the global agricultural sector.