Chicken Disease Classification Based on Inception V3 Algorithm for Data Imbalance

Abstract

In order to supply the world's protein needs, one of the most crucial industries is the poultry business. The problem that often occurs in chicken farms is disease, and this can have a significant impact on the farm. The availability of large enough amounts of data makes it possible to carry out the process of monitoring chicken diseases using deep learning technology for the classification of chicken diseases. With the availability of large enough data, the dataset has a variety of features that cause problems with data clutter. To overcome the problem of data conflict, an oversampling technique is used to increase the sample data from the minority class so that it has the same value as the other majority classes, and the Inception-V3 algorithm is used to classify chicken diseases based on fecal images. The total number of data used was 8067, which were broken down into the following four categories: Healthy, Salmonella, Coccidiosis, and Newcastle disease. Data balancing was done using oversampling to get the total data to 10500 before the evaluation process was started. The data was distributed by splitting it by 80% of the data will be used for training, 10% for data validation, and 10% for testing. The results of the test, which employed Inception V3 without oversampling, produced the highest possible score of 94.05%.