Recognition of Assembly Parts by Convolutional Neural Networks Kamil Židek (&) , Alexander Hosovsky , Jan Piteľ , and Slavomír Bednár Faculty of Manufacturing Technologies with a seat in Presov, Department of Industrial Engineering and Informatics, Technical University of Kosice, Bayerova 1, Presov, Slovak Republic {kamil.zidek,alexander.hosovsky,jan.pitel, slavomir.bednar}@tuke.sk Abstract. The paper describes the experiments with the use of deep neural networks (CNN) for robust identification of assembly parts (screws, nuts) and assembly features (holes), to speed up any assembly process with augmented reality application. The simple image processing tasks with static camera and recognized parts can be handled by standard image processing algorithms (threshold, Hough line/circle detection and contour detection), but the aug- mented reality devices require dynamic recognition of features detected in various distances and angles. The problem can be solved by deep learning CNN which is robust to orientation, scale and in cases when element is not fully visible. We tested two pretrained CNN models Mobilenet V1 and SSD Fast RCNN Inception V2 SSD extension have been tested to detect exact position. The results obtained were very promising in comparison to standard image processing techniques. Keywords: Deep learning Á Object recognition Á Augmented reality 1 Introduction to Augmented Reality and Deep Learning in Industrial Tasks Today’s consumer industry is made up of a large number of products and their possible configurations, which is a direct response to the ever-growing demand of customers. It is generally known that traditional assembly lines are synchronous. This means that the flow of material and work is predefined depending on the customer ’s order. The assembly steps are continuously delegated and performed on each workstation. At the same time all the workstations on the line are synchronized. With the arrival of the Industry 4.0 concept and deployment of its supporting technologies (digital twin, RFID technology, virtual and augmented reality, big data, deep learning), there is also the necessity of changing and deploying asynchronous assembly lines. Applications of such lines can be found in several areas of industry: consumer electronics, furniture, clothing, and automotive production. Because of the variation in production, it is almost unnecessary to co-operate with the machine during the assembly process. For the Industry 4.0 concept, cooperative robots were defined as the main element suitable © Springer Nature Switzerland AG 2019 S. Hloch et al. (Eds.): ICMEM 2018, LNME, pp. 281–289, 2019. https://doi.org/10.1007/978-3-319-99353-9_30