LG Electronics has introduced a highly-integrated LG8111 AI Chip together with a corresponding Eris Reference Board for edge AI Processing.
The LG8111 AI chip for on-device AI inference is equipped with an LG-specific AI processor that can copy the neural network of the human brain and process the deep learning algorithm efficiently. The SoC supports hardware processing in AI capabilities like video, voice, and control intelligence.
Since the chip has a low power low latency feature that enhances its self learning capacity, products built with the chip can implement On-Device AI learned and inferred by themselves. The chip also supports AWS IoT Greengrass and the integration of the Eris Reference board with AWS IoT Greengrass allows for easier deployment of a variety of applications and solutions.
Features of the LG8111 AI SoC:
- LG Neural Engine
- LG proprietary AI Accelerator
- Inference/Training engine
- Support for TensorFlow, TensorFlow Lite, Caffe
- Quad-Core CPU
- ARM Cortex A53 cores clocked at a frequency of 1.0 GHz
- 32KB L1 cache and a 1MB L2 cache
- NEON, FPU, and Cryptography extension.
- Vision Accelerator
- Dedicated vision for DSP which enables feature extraction and image warping.
- Camera Engine:
- 5M + 5M dual ISP pipeline
- Parallel interface
- Noise reduction
- Lens distortion correction
- High Dynamic Range (HDR) pre-processing
- Video Encoder:
- Support for full HD video H.264 encoding
- Audio Engine:
- “Always-on” voice recognition
- Voice Activity Detector with dedicated Audio DSP
The LG LG8111 AI development board on the other hand, is a cost-effective machine learning device development platform for applications such as control intelligence, image classification/recognition, speech recognition, and more. It can provide improved sensor data collection/analysis as well as machine learning inference performance. The board will also find its application in areas like home automation, industrial automation, lighting, etc.