The current number of coronavirus (COVID-19) infections in Indonesia becomes more and more worrying. According to data on June 11, 2020, the number of infected people in Indonesia has reached 35,295 people. With these consequences, it is considered very important to immediately identify infection in order to stop or minimize the spread of the disease. There have been several ways to detect and diagnose COVID-19, one of which is using X-ray images. This paper examines the use of in-depth features and methods to process two-dimensional data from patients' X-ray images. Convolutional Neural Network (CNN) is a development of Multi-Layer Perceptron (MLP), which is specifically designed to process two-dimensional data or image data. The deep features of the fully connected layer CNN model are extracted and can be immediately classified without the need for any additional techniques. CNN method is used because of its good performance for large datasets that will be used for training and testing. In the classification process, the dataset contains 160 x-ray images and consists of two categories, COVID-19 and normal, that represents a positive or negative classification of Covid-19 infection to a patient. To get the best accuracy of the classification model, the author changed several parameters on CNN, such as the distribution of the dataset and the number of epochs. From the nine models tested, model number 5 and 8 with a dataset ratio of 70:30 and epoch number 30 and 40 respectively, resulted in the best accuracy of 97.91%.