Enhancing Green Bean Anthracnose Severity Detection via Integrated CNN-LSTM Models

Anthracnose, a disease resulting from Colletotrichum fungal species, is one of the most dangerous threats to green pea (Phaseolus vulgaris L.) production worldwide at present. Spotting, recognizing, and taking appropriate actions on anthracnose severity levels would be of great value to putting into practice management strategies for disease control. In this study, an innovative and one-of-a-kind hybrid model of convolutional neural network (CNN)- Longshort-term memory (LSTM) is proposed to detect the severity of anthracnose disease in green bean plants. Images of plants infected and already labeled in the range of 0 (without visible symptoms) to 3 (serious symptoms) were collected in the digital format. preprocessing was done on the dataset that included resizing, normalization, and extra images for better result effectiveness. In parallel, the CNN-LSTM architecture recognized the spatial characteristics of the images with the CNN part and relied on the LSTM to handle the temporal relations between the video sequences. The experimental results showed the ability of the model to precisely detect intensity levels leading to $83 \%$ accuracy and an F1 score of 0.88. Comparison tests against established models further brought out the strong points of a new model. CNN-LSTM model presents a worthy solution in the automatic recognition of specific diseases in farming and may be utilized in the science of precision agriculture and crop management.