From February 27 until March 1, 2018, MVTec Software GmbH, the leading provider of machine vision software, demonstrates the robustness, speed, and reliability of machine vision in embedded systems at embedded world 2018 in Nuremberg. At booth 203 in hall 4, visitors can experience the performance of the machine vision software products HALCON and MERLIC in embedded applications. Presentations by MVTec staff in the Exhibitors’ Forum and at the embedded world Conference give deeper insights into capabilities and theory of embedded vision.
Embedded vision and deep learning demonstrations, expert lectures
HALCON 17.12 Progress, the newest version of HALCON, lets users profit from the power of deep learning. Here, pretrained networks based on industrial image data serve as a starting point, so there is no need to train from scratch. A live demonstration at the MVTec booth shows how embedded devices can be used for deep learning, with an NVIDIA Jetson TX2 classifying various objects in real time.
HALCON’s compatibility to many hardware standards is a major advantage of the software. This is demonstrated in a second live demonstration involving four embedded boards and a linear sled. The four boards are each set to solve a vision task, and their respective cameras are aligned to different spots on an analysis object that gets moved horizontally at high speeds. This shows the speed and robustness of HALCON’s performance on embedded hardware.
MVTec MERLIC demonstrates the power of deep-learning-based OCR in a third live demonstration by robustly and quickly reading expiration dates on medication packagings. Here, all processing happens on a smart camera, eliminating the necessity of an industrial PC.
Furthermore, experts from MVTec round off the program by holding presentations at the embedded world Conference on the topic “software development for embedded vision” as well as at the Exhibitors’ Forum on the topic “deep learning for image classification”.