Deep Learning Edge Smart Camera for Inference Applications

Smart Solutions

Deep Learning Edge Smart Camera for Inference Applications

The smart camera Icam-7000 series is the first deep learning edge smart camera specially developed for inference applications.

The smart camera Icam-7000 is pre-installed with EzBuilder, a graphical user interface-based application software which supports various machine vision tools. (Image: Advantech Europe B.V.)

The small smart camera series includes monochrome and color models, providing up to 5MP resolution and featuring a global shutter sensor. The camera’s high-performance multi-core CPU+FPGA computing system increases computing efficiency, while reducing power consumption. To support industrial applications in harsh environments, the small IP67-rated smart cameras can be integrated easily with existing infrastructure and equipment. The Icam-7000 is pre-installed with EzBuilder (Easy Builder) – ensuring suitability for a wide range of applications. The graphical user interface-based application software supports various machine vision tools. With the software, users only need to purchase licenses for the specific machine vision functions they require, such as identification functions for traceability applications or alignment functions for vision guidance and pattern matching applications. The flowchart-based design of the graphical user interface simplifies task execution to three simple steps: eliminating the need for complex programming. This means that even users without programming skills can complete the software and application setup. Moreover, the software supports web-based remote monitoring via Ethernet, allowing users to remotely access/manage cameras from any browser. The series of smart cameras also supports the Intel OpenVino deep learning toolkit, allowing smart cameras to rapidly adapt to new inspection criteria without system reconfiguration. Users only need to upload sample images to the training server, run the training process using the Intel deep learning suite, and finally export a trained model for inference by the camera. After inference, the smart camera is ready to be deployed for object recognition. Advantech’s Rick de Vries, explains: “Traditional OCR on its own can struggle to effectively process distortions, as well as different fonts, sizes and languages. We’ve found a smart and efficient solution through Deep Learning. Our OCR uses image labelling, training and inferencing processes to give businesses greater reliability on the factory floor.

Advantech Europe BV