Chicken Disease Detection Based on Fases Image Using EfficientNetV2L Model

Livestock farming requires technological innovation to increase productivity and efficiency. Chickens are a livestock animal with good market prospects. However, not all farmers understand about chicken diseases and signs of sickness. Detection of chicken diseases can be done through various methods, one of which is by looking at the shape of the chicken's feces. Images in feces can be detected using machine learning. Convolutional Neural Networks (CNN) are used to speed up disease prediction. Transfer learning is used to leverage knowledge that has been learned by previous models. In this study, we propose our own CNN architecture model and present research by building a new model to detect and classify diseases in chickens through their feces. The model training process is carried out by inputting training data and validation data, the number of epochs, and the created checkpointer object. The hyperparameter tuning stage is carried out to increase the accuracy rate of the model. The research is conducted by testing datasets obtained from the Kaggle repository which has images of coccidiosis, salmonella, Newcastle, and healthy feces. The results of the study show that our proposed model only achieves an accuracy rate of 93%, while the best accuracy rate in the study is achieved by using the EfficientNerV2L model with the RMSProp optimizer, which is 97%.