LILIN Launches AI Camera carries user pre-trained models to perform inference without PC
LILIN, a leading AI video solution provider, launches Self+AI Cloud Service, providing modular AI plug-ins that allow users to convert pre-trained object models into a common package for LILIN front-end AI cameras to use with behaviors for creating various AI applications. The image inference process requires no PC, and the results can then be integrated via Python or C# SDKs, or sent through hardware by digital output, which performs a "one-stop" AIoT edge solution.
AI image inference was difficult to implement in edge applications. It required a PC host to perform the computation, which increases the complexity of the wiring and architecture, and the AI models are usually limited or not compatible. Plus lots of cameras are lack behaviors, to build an application was very difficult.
The LILIN Edge Computing Aida 7 Series cameras are powered by the Ambarella CV Series SoC with a 4-core ARM® Cortex®-A53 CPU that can operate at 1.0 GHz, with a high-speed AI GPU chip, and 1 GB of eMMC flash memory, the performance can beat most of AI USB sticks and stay with mainstream PCs.
The Self+AI Cloud, developed by the LILIN AI team, converts the Tiny Yolo AI model into a compatible format running by the Ambarella CV Series SoC and packaged in an LPKG (LILIN Package) for distribution to compatible cameras without writing a single line of code. With the rich behavior settings of LILIN cameras, the threshold for developing AI applications becomes extremely low.
A typical case is the implementation of ship-tracking. The Ship and Ocean Industries R&D Center in Taiwan uses a LILIN PTZ camera which carries NEMA 4X, IK10, IP68 certified, and infrared operation to identify vessels and track them.
Dr. Zhou Xianguang, Chief Executive Officer of the Ship and Ocean Industries R&D Center, said "The LILIN AI platform and its cameras are innovative, allowing our AI algorithms to be smoothly integrated. The AI conversion cloud developed by LILIN also significantly shortens the development phase for our team members to deploy solutions. Edge computing cameras definitely will play an important role in future AI applications.
"The issue of harbor surveillance and ships management requires cameras on the fleet and shore to assist navigation. LILIN provided a complete platform to quickly deploy our algorithm to the scenario." Director Liu Jianhong of the Ship and Ocean Industries R&D Center, also mentioned, "The harbor environment is complex and harsh, and ships are not suitable for large computing hosts due to salt spray, power supply, and other rugged issues. LILIN's NEMA 4x AI camera needs no PC to work, hence fitting marine applications.
Mr. Hu, the CIO of LILIN also added, "From Intel OpenVINO, NVIDIA CUDA to Ambarella, LILIN's AI framework includes front-end and back-end, and it can be ported across platforms. We expect the LILIN Self+AI platform to become an 'algorithm deployment accelerator' for AI start-ups, achieving a multi-win situation in the AI ecosystem."
LILIN also offers Python and C# code on Github for any engineer familiar with these languages to immediately fork and develop.
Key Specifications of LILIN 7 Series Edge Computing Cameras
● 4K high-definition image quality, no image details can be hidden
● Smart H.265 video compression with low bandwidth consumption
● Nighttime ultra low light color mode for more accurate AI recognition
● HDR Super Wide Dynamic 120 dB, suitable for backlight environment
● Ultra-fast shutter in 1/20000s, high success rate of license plate recognition
White light for parking lots
P-IRIS model for morning and evening scenes