Shoe prints belong among the most common types of evidence at crime scenes. To determine the brand or manufacturer of a shoe, it must be matched with known collection of shoe print samples, or with shoe sole pictures. Comparing the shoe print to the shoe sole picture is difficult task, because of different nature of data. Therefore it is more convenient to compare shoe sole features instead of raw image data. In this paper we focus on detection of the shoe sole features in picture of shoe sole using Deep Neural Network.