About six months after the release of MERLIC 5, an enhanced feature set is now available that provides even better usability. In particular, the range of functions of the deep learning AI technology was expanded. In addition to the anomaly detection and classification features that were already included in MERLIC 5, Deep OCR has now been added. MVTec has also optimized hardware compatibility in MERLIC 5.1, which greatly accelerates deep learning applications.
“MERLIC 5.1 is inspired by our belief that even users without any in-depth programming knowledge should be able to benefit from machine vision. MERLIC 5 was already a tremendous success and raised the software to a whole new level. With version 5.1, we’re continuing along the same path – for example, by relying even more on innovative deep learning technologies and integrating new functions so that more users can take advantage of our easy-to-use machine vision software,” explains Christoph Wagner, Technical Product Manager MERLIC at MVTec.
Robust character recognition with Deep OCR
MERLIC 5.1 includes a new tool for optical character recognition that is based on HALCON’s Deep OCR technology. Compared to other algorithms, this holistic deep-learning-based approach can localize characters much more robustly, regardless of their orientation, font type, polarity, and it requires significantly less parameter tuning. Recognition performance is further increased by the automatic grouping of characters. This allows the identification of whole words and thus reduces the chance of misinterpretation of similar-looking characters. This new tool does not require the Deep Learning Add-On and is therefore already included in the entry level package MERLIC 5.1 S.
Support for AI accelerator hardware
MERLIC 5.1 also includes Artificial Intelligence Acceleration Interface (AI²) plug-ins for the NVIDIA® TensorRT™ SDK and the Intel® Distribution of OpenVINO™ toolkit. This enables users to benefit from compatible AI accelerator hardware – quickly and conveniently. MERLIC tools using deep learning functionality can thus achieve significantly faster inference times when paired with compatible hardware like NVIDIA GPUs or Intel processors including GPUs, CPUs and VPUs like the Intel® Movidius™ Neural Compute Stick. By expanding the range of supported Intel devices, customers now enjoy even more flexibility in their choice of hardware. By adding support for AI², MERLIC will also benefit from any future plug-ins that integrate new accelerator hardware.
Image Source Manager improvements
And finally, selecting and setting the correct camera parameters within the Image Source Manager (ISM) is further simplified in MERLIC 5.1. When a camera is set up, the required parameters can be found much faster and with greater ease. The search for parameters can now be carried out via the ISM GUI, similar to the search function in the Tool Library. In addition, it is possible to filter for persistent parameters that are stored in the camera, even if the camera is switched off. Furthermore, configurations and image sources can now be renamed in MERLIC RTE Setup. Last but not least, the connection status of cameras that support this feature is now displayed.