Meat is one of the food ingredients needed by humans. The price of pork is cheaper than beef, which has led to the practice of mixing beef with pork for the purpose of making big profits. In plain view, the difference between beef and pork is not striking, so it is difficult for ordinary people to distinguish between them. In terms of color, pork is paler than beef. In terms of texture, beef is stiffer and tougher than pork. In terms of fiber, beef is clearer than pork, so we need a system that can identify the two types of meat. This study uses the Convolutional Neural Network (CNN) algorithm with the ResNet-50 architecture with 3 types of optimizers such as Stochastic Gradient Descent (SGD), Adam, and RMSprop. The dataset used for training first goes through 2 stages of preprocessing, namely cropping and resizing. The results of the study show that the SGD optimizer can outperform the Adam and RMSprop optimizers with 97.83% accuracy, 97% precision, 97% recall, and 97% f1 score with batch size 32, learning rate 0.01, and epoch 50.

Beef is an example of an animal protein-rich food. The consumption of meat in Indonesia is increasing year after year, in tandem with the country's growing population. Many traders purposefully combine beef and pork in order to maximize profits. With the naked eye, it's difficult to tell the difference between pork and beef. In Muslim-majority countries, the assurance of halal meat is crucial. This study uses Deep Learning with the Convolutional Neural Network (CNN) method and ResNet-50 with data augmentation to classify images of beef and pork. The original meat picture databases contain 457 images, however following the data augmentation process, there are 2742 images in total, divided into three classes. The distribution of training and test data is 90 percent:10 percent in the comparison test scenario between the two original data schemes and supplemented data. With an average of 87.64 % accuracy, 87.59 % recall, and 90.90 % precision, the Confusion Matrix is the best classification performance model. There was no evidence of overfitting based on observations from the visualization of the training and testing process.

Cases of mixing beef and pork are still happening today. The increasing demand for beef causes many traders to mix meat to gain more profit. Distinguishing beef and pork can be done by sight and smell, but still has weaknesses. This study uses Deep Learning method for image classification with Convolutional Neural Network architecture EfficientNet-B0. The amount of data is 3,000 images which are divided into 3 classes, beef, pork, and mixed meat. This study uses original image data and image data of Contrast Limited Adaptive Histogram Equalization. The data is divided by the ratio of training data and test data of 80:20. The results of testing the model with the confusion matrix show the highest classification performance with 95.17% accuracy, 92.72% precision, 95.5% recall, and 94.09% f1 score, in the original image data with the use of neurons in the first dense amounting to 256, 32 batch size, 0.01 learning rate, and Adam's optimizer

Abstrak - Kasus kecurangan pedagang mencampur daging sapi dengan daging babi masih terjadi hingga saat ini. Membedakan daging sapi dan babi dapat dilakukan dengan mengamati secara langsung satu persatu, tetapi hal ini dapat dilakukan oleh para ahli, Tetapi secara kasat mata masih sulit membedakannya. Perilaku pedagang seperti ini sangat merugikan konsumen khususnya pemeluk agama Islam karena berkaitan dengan makanan yang halal atau haram. Pada penelitain ini menggunakan metode Deep Learning untuk klasifikasi citra dengan Convolutional Neural Network (CNN) arsitektur ResNet-50. Jumlah data sebanyak 457 citra yang terbagi menjadi 3 kelas, yaitu daging babi, daging oplosan dan daging sapi. Setiap kelas memiliki ukuran gambar yang sama yaitu 300 x 300 pixel. Pembagian data menggunakan split data dengan perbandingan 70% data uji : 30% data uji, 80% data latih : 20% data uji, dan 90% data latih : 10% data uji. Hasil dari pengujian model dengan Confusion Matrix menunjukkan performa klasifikasi tertinggi dengan 100% accuracy, 100% precision, dan 100% recall, pada data citra asli dengan penggunaan batch size 32, 0.001 learning rate, epoch 75 dan split data 90% : 10%.Kata kunci: Convolutional Neural Network, Daging Babi dan Sapi, Deep Learning, Klasifikasi Citra, ResNet  Abstract - Traders mixing beef and pork are still committing fraud today. Although professionals can discern between beef and pork by watching them one by one, it is still impossible to do so with the naked eye. This kind of behavior is very detrimental to consumers, especially Muslims because it is related to halal or haram food. This research uses Deep Learning method to classify images with Convolutional Neural Network (CNN) ResNet-50 architecture. The number of data is 457 images which are divided into 3 classes, namely pork, mixed meat and beef. Each class has the same image size, which is 300 x 300 pixels. data distribution using split data with a comparison of 70% training data: 30% test data, 80% training data: 20% test data, and 90% training data: 10% test data. The results of model testing using the Confusion Matrix show the highest classification performance with 100% accuracy, 100% precision, and 100% recall, on the original image data using batch size 32, 0.001 learning rate, epoch 75 and split data 90%: 10%..Kata kunci: Convolutional Neural Networ, Deep Learning, Image Classification, Pork and Beef, ResNet

The demand for meat began to increase rapidly, causing drastic price changes and causing the existence of scammers to inflate the price of meat to get big profits by mixing beef and pork. Few consumers are aware of the mixing of meat, to distinguish between beef and pork can be seen in terms of color and texture, but this theory still has weaknesses. This research uses the Deep Learning method, namely Convolutional Neural Network with Local Binary Pattern texture extraction feature and AlexNet architecture for meat classification. The research conducted stated that the accuracy of the meat image classification can be measured using various parameters and optimizers. The highest accuracy results obtained from this study were 68.6% accuracy, 62% precision, 57.6% recall, and 59% f1-score using the Stochastic Gradient Descent (SGD) optimizer, 0.01 learning rate, 32 batch size, and 0.9 momentum. Compared to the original dataset, the accuracy of the LBP dataset type is still below the original dataset with the results obtained from the accuracy of the original dataset are 84.1% accuracy, 78.6% precision, 79% recall, and 79% f1-score using the RMSprop optimizer, 0 .0001 learning rate, 32 batch sizes, and momentum So it can be concluded that the AlexNet architecture by setting the existing parameter values can increase the accuracy value.