| dc.description.abstract | Adulteration of feed ingredients is one of the main problems in ensuring quality. This study aims to identify the level of fish meal forgery using rice bran based on image analysis with a convolutional neural network algorithm. The number of datasets in this study was 3200 images which were divided into 2400 images for training data and 800 images for test data. The forgery treatment consists of P0=100% fish meal, P1= 90% fish meal+10% rice bran, P2=80% fish meal+20% rice bran, P3= 70% fish meal+30% rice bran and P4=100% rice bran. In the CNN algorithm, the image is processed through the input layer, feature extraction layer and fully connected. The results of this study obtained a training data accuracy value of 100% and a validation data accuracy value of 100%. In testing through the confusion matrix table, the results of 100% accuracy, 99% precision and 100% recal using 20 epochs were obtained. Based on this study, it was concluded that testing the counterfeiting of fish meal and rice bran using CNN can provide quite good and optimal results. Keyswords: adulteration analysis of feed ingredients, convolutional neural network, fish meal, image analysis, rice brain | id |
