Their calculation requires specialized hardware and a high bandwidth for data communication. In particular, the provision and processing of image data from connected cameras poses challenges for controllers used in industrial environments. They need to be space-saving, robust and powerful, while at the same time requiring little maintenance and being equipped with the essential interfaces. With the DLAP-301-Nano, Acceed is now announcing a new AI platform with an integrated Jetson nano engine that is tailored to these requirements.

Equipped with the Jetson Nano module from Nvidia, the new fanless DLAP-301-Nano presented at Acceed is the ideal platform when it comes to getting started with AI applications based on image data. The module has a 64-bit quad-core ARM Cortex-A57 CPU and an integrated 128-CUDA core graphics processor (GPU) with Maxwell architecture. In combination with 4 GB of 64-bit LPDDR4 memory, the module achieves 472 GFLOPS of computing power. Nvidia specifies a memory throughput of 25.6 GBit/s. The power consumption is only 5 to 10 W, depending on the activated mode.

Eight RJ45 Ethernet interfaces with PoE are used to connect cameras at the rear of the industrial housing, which is only 21 cm wide. Two serial DB-9 sockets, three USB ports and a further GbE interface are also available for data communication. On-site graphics output is possible in high resolution via the HDMI 2.0 port on the front. The bay, which is also accessible on the front, accommodates a 2.5" SATA SSD for local data storage. This means that the DLAP-301-Nano can also be used as an autonomous NVR (network video recorder), which can also stream image data to the network if required.

With the Jetson Nano module, the DLAP-301-Nano enables sophisticated computer vision in real time and inferencing for many complex DNN (deep neural network) models. This could be used, for example, to control multisensory autonomous robots or to integrate IoT devices with intelligent edge processing. With the help of ML frameworks, it should also be possible to train neural networks directly.

The DLAP-301-Nano has been developed for industrial use under demanding conditions and allows operation in the extended temperature range from -20 to 70 °C. The power supply is 12 V DC. The robust full metal chassis is very compact at 210 x 170 x 55 mm.

Adlink's AI platforms are available in Germany via the official sales partner Acceed. The focus is on Industrial Ethernet, embedded AI, ready-made IPCs for automation, building management, logistics, machine vision and research and development.