Advances in Industry 4.0, IoT and mobile systems have led to an increase in the number of malware threats that target these systems. The research shows that classification via the use of computer vision and machine learning methods over byte-level images extracted from malware files could be an effective static solution. In this study, in order to detect malware, we have employed various contemporary convolutional neural networks (Resnet, Inception, DenseNet, VGG, AlexNet) that have proven success in image classification problem and compared their predictive performance along with duration of model production and inference. In addition, a novel malware data set involving 8750 training and 3644 test instances over 25 different classes was proposed and used. As a result of the experiments carried out with 3-channel (RGB) images obtained, the highest success in terms of accuracy was determined as 97.48% by using DenseNet networks.